这是一个自动调整超参数的功能演示，
首先该程序有一个bug，可能在经过很长一段时间的训练之后，最后一步报错了，
Traceback (most recent call last):
File "mnist_sklearn_wrapper.py", line 96, in <module>
validator.fit(x_train, y_train)
File "C:ProgramDataMiniconda3libsite-packagessklearnmodel_selection\_search.py", line 736, in fit
**self.best_params_))
File "C:ProgramDataMiniconda3libsite-packagessklearnase.py", line 82, in clone
(estimator, name))
RuntimeError: Cannot clone object <keras.wrappers.scikit_learn.KerasClassifier object at 0x000001300883B848>, as the constructor either does not set or modifies parameter dense_layer_sizes我看官方有人提交了一个pull request，
https://github.com/keras-team/keras/pull/13598/commits/c735ab5b89bbf935075c84aab3437468e1fe8245
修改后可以正常运行，但该修改的集成测试却没有通过，所以可能不算是一个完美修复
不过可以临时按照它的方式进行修改，在测试完后再改回来
GridSearchCV 提供了几组候选参数，
len(dense_layer_sizes)=4
len(epochs)=2
len(filters)=1
len(kernel_size)=1
len(pool_size)=1排列组合一下，总共是4*2*1*1*1，总共八种情况，也就是说可以组成 8 种神经网络训练方式，来比较一下，哪种排列组合的训练效果是最好的；每种训练方式会重复五次，也就是总共会训练 40 次；
下面附完整训练日志：
Using TensorFlow backend.
2020-04-03 14:52:24.644605: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll
dense_layer_sizes [32]
filters 8
kernel_size 3
pool_size 2
2020-04-03 14:52:27.891345: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library nvcuda.dll
2020-04-03 14:52:28.988239: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1555] Found device 0 with properties:
pciBusID: 0000:01:00.0 name: Quadro RTX 3000 computeCapability: 7.5
coreClock: 1.38GHz coreCount: 30 deviceMemorySize: 6.00GiB deviceMemoryBand 312.97GiB/s
2020-04-03 14:52:28.994920: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll
2020-04-03 14:52:29.003690: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cublas64_10.dll
2020-04-03 14:52:29.013001: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cufft64_10.dll
2020-04-03 14:52:29.018018: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library curand64_10.dll
2020-04-03 14:52:29.025735: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusolver64_10.dll
2020-04-03 14:52:29.034548: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusparse64_10.dll
2020-04-03 14:52:29.047961: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudnn64_7.dll
2020-04-03 14:52:29.052178: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1697] Adding visible gpu devices: 0
2020-04-03 14:52:29.054762: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2
2020-04-03 14:52:29.067432: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1555] Found device 0 with properties:
pciBusID: 0000:01:00.0 name: Quadro RTX 3000 computeCapability: 7.5
coreClock: 1.38GHz coreCount: 30 deviceMemorySize: 6.00GiB deviceMemoryBand 312.97GiB/s
2020-04-03 14:52:29.074938: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll
2020-04-03 14:52:29.080278: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cublas64_10.dll
2020-04-03 14:52:29.082466: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cufft64_10.dll
2020-04-03 14:52:29.085739: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library curand64_10.dll
2020-04-03 14:52:29.089171: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusolver64_10.dll
2020-04-03 14:52:29.091364: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cusparse64_10.dll
2020-04-03 14:52:29.095283: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudnn64_7.dll
2020-04-03 14:52:29.097662: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1697] Adding visible gpu devices: 0
2020-04-03 14:52:29.638757: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1096] Device interconnect StreamExecutor with strength 1 edge matrix:
2020-04-03 14:52:29.642200: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1102] 0
2020-04-03 14:52:29.644956: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] 0: N
2020-04-03 14:52:29.647308: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1241] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 4604 MB memory) -> physical GPU (device: 0, name: Quadro RTX 3000, pci bus id: 0000:01:00.0, compute capability: 7.5)
Model: "sequential_1"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_1 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_1 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_2 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_2 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_1 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_1 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_1 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_1 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_3 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_2 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_2 (Dense) (None, 10) 330
________________________________________________________________________________
activation_4 (Activation) (None, 10) 0
================================================================================
Total params: 37,890
Trainable params: 37,890
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
2020-04-03 14:52:30.684354: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cublas64_10.dll
2020-04-03 14:52:30.950919: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudnn64_7.dll
2020-04-03 14:52:32.029379: W tensorflow/stream_executor/gpu/redzone_allocator.cc:312] Internal: Invoking GPU asm compilation is supported on Cuda non-Windows platforms only
Relying on driver to perform ptx compilation. This message will be only logged once.
48000/48000 [==============================] - 6s 135us/step - loss: 0.6358 - accuracy: 0.7876
Epoch 2/3
48000/48000 [==============================] - 5s 100us/step - loss: 0.3572 - accuracy: 0.8862
Epoch 3/3
48000/48000 [==============================] - 5s 100us/step - loss: 0.2977 - accuracy: 0.9071
dense_layer_sizes [32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_2"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_3 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_5 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_4 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_6 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_2 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_3 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_2 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_3 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_7 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_4 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_4 (Dense) (None, 10) 330
________________________________________________________________________________
activation_8 (Activation) (None, 10) 0
================================================================================
Total params: 37,890
Trainable params: 37,890
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 104us/step - loss: 0.6602 - accuracy: 0.7871
Epoch 2/3
48000/48000 [==============================] - 5s 101us/step - loss: 0.3518 - accuracy: 0.8912
Epoch 3/3
48000/48000 [==============================] - 5s 100us/step - loss: 0.2985 - accuracy: 0.9089
dense_layer_sizes [32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_3"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_5 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_9 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_6 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_10 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_3 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_5 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_3 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_5 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_11 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_6 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_6 (Dense) (None, 10) 330
________________________________________________________________________________
activation_12 (Activation) (None, 10) 0
================================================================================
Total params: 37,890
Trainable params: 37,890
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 96us/step - loss: 0.6220 - accuracy: 0.7967
Epoch 2/3
48000/48000 [==============================] - 5s 101us/step - loss: 0.3455 - accuracy: 0.8933
Epoch 3/3
48000/48000 [==============================] - 4s 93us/step - loss: 0.2835 - accuracy: 0.9113
dense_layer_sizes [32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_4"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_7 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_13 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_8 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_14 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_4 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_7 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_4 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_7 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_15 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_8 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_8 (Dense) (None, 10) 330
________________________________________________________________________________
activation_16 (Activation) (None, 10) 0
================================================================================
Total params: 37,890
Trainable params: 37,890
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 108us/step - loss: 0.6077 - accuracy: 0.8029
Epoch 2/3
48000/48000 [==============================] - 5s 104us/step - loss: 0.3222 - accuracy: 0.9013
Epoch 3/3
48000/48000 [==============================] - 5s 101us/step - loss: 0.2716 - accuracy: 0.9163
dense_layer_sizes [32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_5"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_9 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_17 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_10 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_18 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_5 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_9 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_5 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_9 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_19 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_10 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_10 (Dense) (None, 10) 330
________________________________________________________________________________
activation_20 (Activation) (None, 10) 0
================================================================================
Total params: 37,890
Trainable params: 37,890
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 96us/step - loss: 0.5510 - accuracy: 0.8200
Epoch 2/3
48000/48000 [==============================] - 5s 101us/step - loss: 0.3218 - accuracy: 0.8961
Epoch 3/3
48000/48000 [==============================] - 5s 104us/step - loss: 0.2708 - accuracy: 0.9138
dense_layer_sizes [32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_6"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_11 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_21 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_12 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_22 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_6 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_11 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_6 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_11 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_23 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_12 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_12 (Dense) (None, 10) 330
________________________________________________________________________________
activation_24 (Activation) (None, 10) 0
================================================================================
Total params: 37,890
Trainable params: 37,890
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 105us/step - loss: 0.6296 - accuracy: 0.7956
Epoch 2/6
48000/48000 [==============================] - 4s 93us/step - loss: 0.3656 - accuracy: 0.8840
Epoch 3/6
48000/48000 [==============================] - 5s 95us/step - loss: 0.3040 - accuracy: 0.9047
Epoch 4/6
48000/48000 [==============================] - 5s 94us/step - loss: 0.2726 - accuracy: 0.9158
Epoch 5/6
48000/48000 [==============================] - 4s 92us/step - loss: 0.2573 - accuracy: 0.9214
Epoch 6/6
48000/48000 [==============================] - 4s 93us/step - loss: 0.2341 - accuracy: 0.9281
dense_layer_sizes [32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_7"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_13 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_25 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_14 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_26 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_7 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_13 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_7 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_13 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_27 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_14 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_14 (Dense) (None, 10) 330
________________________________________________________________________________
activation_28 (Activation) (None, 10) 0
================================================================================
Total params: 37,890
Trainable params: 37,890
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 103us/step - loss: 0.6525 - accuracy: 0.7845
Epoch 2/6
48000/48000 [==============================] - 5s 104us/step - loss: 0.3825 - accuracy: 0.8804
Epoch 3/6
48000/48000 [==============================] - 5s 106us/step - loss: 0.3172 - accuracy: 0.9012
Epoch 4/6
48000/48000 [==============================] - 5s 103us/step - loss: 0.2687 - accuracy: 0.9178
Epoch 5/6
48000/48000 [==============================] - 5s 99us/step - loss: 0.2373 - accuracy: 0.9275
Epoch 6/6
48000/48000 [==============================] - 5s 101us/step - loss: 0.2251 - accuracy: 0.9293
dense_layer_sizes [32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_8"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_15 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_29 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_16 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_30 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_8 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_15 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_8 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_15 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_31 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_16 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_16 (Dense) (None, 10) 330
________________________________________________________________________________
activation_32 (Activation) (None, 10) 0
================================================================================
Total params: 37,890
Trainable params: 37,890
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 99us/step - loss: 0.5917 - accuracy: 0.8095
Epoch 2/6
48000/48000 [==============================] - 4s 89us/step - loss: 0.3464 - accuracy: 0.8930
Epoch 3/6
48000/48000 [==============================] - 5s 95us/step - loss: 0.3023 - accuracy: 0.9087
Epoch 4/6
48000/48000 [==============================] - 4s 89us/step - loss: 0.2709 - accuracy: 0.9164
Epoch 5/6
48000/48000 [==============================] - 4s 91us/step - loss: 0.2518 - accuracy: 0.9237
Epoch 6/6
48000/48000 [==============================] - 4s 90us/step - loss: 0.2447 - accuracy: 0.9250
dense_layer_sizes [32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_9"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_17 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_33 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_18 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_34 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_9 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_17 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_9 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_17 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_35 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_18 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_18 (Dense) (None, 10) 330
________________________________________________________________________________
activation_36 (Activation) (None, 10) 0
================================================================================
Total params: 37,890
Trainable params: 37,890
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 98us/step - loss: 0.5954 - accuracy: 0.8045
Epoch 2/6
48000/48000 [==============================] - 4s 93us/step - loss: 0.3322 - accuracy: 0.8965
Epoch 3/6
48000/48000 [==============================] - 4s 92us/step - loss: 0.2954 - accuracy: 0.9091
Epoch 4/6
48000/48000 [==============================] - 5s 96us/step - loss: 0.2624 - accuracy: 0.9190
Epoch 5/6
48000/48000 [==============================] - 5s 101us/step - loss: 0.2439 - accuracy: 0.9249
Epoch 6/6
48000/48000 [==============================] - 5s 98us/step - loss: 0.2308 - accuracy: 0.9290
dense_layer_sizes [32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_10"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_19 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_37 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_20 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_38 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_10 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_19 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_10 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_19 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_39 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_20 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_20 (Dense) (None, 10) 330
________________________________________________________________________________
activation_40 (Activation) (None, 10) 0
================================================================================
Total params: 37,890
Trainable params: 37,890
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 104us/step - loss: 0.5412 - accuracy: 0.8263
Epoch 2/6
48000/48000 [==============================] - 4s 94us/step - loss: 0.3317 - accuracy: 0.8970
Epoch 3/6
48000/48000 [==============================] - 4s 91us/step - loss: 0.2831 - accuracy: 0.9139
Epoch 4/6
48000/48000 [==============================] - 5s 94us/step - loss: 0.2618 - accuracy: 0.9202
Epoch 5/6
48000/48000 [==============================] - 4s 92us/step - loss: 0.2422 - accuracy: 0.9269
Epoch 6/6
48000/48000 [==============================] - 5s 95us/step - loss: 0.2302 - accuracy: 0.9302
dense_layer_sizes [64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_11"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_21 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_41 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_22 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_42 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_11 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_21 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_11 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_21 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_43 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_22 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_22 (Dense) (None, 10) 650
________________________________________________________________________________
activation_44 (Activation) (None, 10) 0
================================================================================
Total params: 75,106
Trainable params: 75,106
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 102us/step - loss: 0.4149 - accuracy: 0.8740
Epoch 2/3
48000/48000 [==============================] - 5s 96us/step - loss: 0.2064 - accuracy: 0.9419
Epoch 3/3
48000/48000 [==============================] - 4s 92us/step - loss: 0.1668 - accuracy: 0.9507
dense_layer_sizes [64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_12"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_23 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_45 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_24 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_46 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_12 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_23 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_12 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_23 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_47 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_24 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_24 (Dense) (None, 10) 650
________________________________________________________________________________
activation_48 (Activation) (None, 10) 0
================================================================================
Total params: 75,106
Trainable params: 75,106
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 96us/step - loss: 0.4150 - accuracy: 0.8691
Epoch 2/3
48000/48000 [==============================] - 5s 95us/step - loss: 0.1982 - accuracy: 0.9417
Epoch 3/3
48000/48000 [==============================] - 4s 93us/step - loss: 0.1593 - accuracy: 0.9532
dense_layer_sizes [64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_13"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_25 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_49 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_26 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_50 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_13 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_25 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_13 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_25 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_51 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_26 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_26 (Dense) (None, 10) 650
________________________________________________________________________________
activation_52 (Activation) (None, 10) 0
================================================================================
Total params: 75,106
Trainable params: 75,106
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 100us/step - loss: 0.4597 - accuracy: 0.8565
Epoch 2/3
48000/48000 [==============================] - 4s 90us/step - loss: 0.2266 - accuracy: 0.9339
Epoch 3/3
48000/48000 [==============================] - 4s 93us/step - loss: 0.1774 - accuracy: 0.9489
dense_layer_sizes [64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_14"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_27 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_53 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_28 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_54 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_14 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_27 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_14 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_27 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_55 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_28 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_28 (Dense) (None, 10) 650
________________________________________________________________________________
activation_56 (Activation) (None, 10) 0
================================================================================
Total params: 75,106
Trainable params: 75,106
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 96us/step - loss: 0.4102 - accuracy: 0.8733
Epoch 2/3
48000/48000 [==============================] - 4s 93us/step - loss: 0.2006 - accuracy: 0.9402
Epoch 3/3
48000/48000 [==============================] - 5s 94us/step - loss: 0.1643 - accuracy: 0.9522
dense_layer_sizes [64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_15"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_29 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_57 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_30 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_58 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_15 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_29 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_15 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_29 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_59 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_30 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_30 (Dense) (None, 10) 650
________________________________________________________________________________
activation_60 (Activation) (None, 10) 0
================================================================================
Total params: 75,106
Trainable params: 75,106
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 108us/step - loss: 0.4029 - accuracy: 0.8747
Epoch 2/3
48000/48000 [==============================] - 5s 98us/step - loss: 0.2007 - accuracy: 0.9412
Epoch 3/3
48000/48000 [==============================] - 5s 97us/step - loss: 0.1701 - accuracy: 0.9515
dense_layer_sizes [64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_16"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_31 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_61 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_32 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_62 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_16 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_31 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_16 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_31 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_63 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_32 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_32 (Dense) (None, 10) 650
________________________________________________________________________________
activation_64 (Activation) (None, 10) 0
================================================================================
Total params: 75,106
Trainable params: 75,106
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 97us/step - loss: 0.4400 - accuracy: 0.8624
Epoch 2/6
48000/48000 [==============================] - 5s 97us/step - loss: 0.2095 - accuracy: 0.9378
Epoch 3/6
48000/48000 [==============================] - 5s 95us/step - loss: 0.1630 - accuracy: 0.9518
Epoch 4/6
48000/48000 [==============================] - 4s 91us/step - loss: 0.1401 - accuracy: 0.9592
Epoch 5/6
48000/48000 [==============================] - 4s 91us/step - loss: 0.1276 - accuracy: 0.9625
Epoch 6/6
48000/48000 [==============================] - 4s 91us/step - loss: 0.1227 - accuracy: 0.9652
dense_layer_sizes [64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_17"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_33 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_65 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_34 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_66 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_17 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_33 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_17 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_33 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_67 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_34 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_34 (Dense) (None, 10) 650
________________________________________________________________________________
activation_68 (Activation) (None, 10) 0
================================================================================
Total params: 75,106
Trainable params: 75,106
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 101us/step - loss: 0.4304 - accuracy: 0.8656
Epoch 2/6
48000/48000 [==============================] - 4s 92us/step - loss: 0.2009 - accuracy: 0.9404
Epoch 3/6
48000/48000 [==============================] - 4s 92us/step - loss: 0.1646 - accuracy: 0.9520
Epoch 4/6
48000/48000 [==============================] - 5s 97us/step - loss: 0.1440 - accuracy: 0.9594
Epoch 5/6
48000/48000 [==============================] - 5s 100us/step - loss: 0.1280 - accuracy: 0.9622
Epoch 6/6
48000/48000 [==============================] - 5s 100us/step - loss: 0.1211 - accuracy: 0.9660
dense_layer_sizes [64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_18"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_35 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_69 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_36 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_70 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_18 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_35 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_18 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_35 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_71 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_36 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_36 (Dense) (None, 10) 650
________________________________________________________________________________
activation_72 (Activation) (None, 10) 0
================================================================================
Total params: 75,106
Trainable params: 75,106
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 96us/step - loss: 0.4330 - accuracy: 0.8669
Epoch 2/6
48000/48000 [==============================] - 5s 94us/step - loss: 0.2064 - accuracy: 0.9399
Epoch 3/6
48000/48000 [==============================] - 5s 95us/step - loss: 0.1694 - accuracy: 0.9508
Epoch 4/6
48000/48000 [==============================] - 4s 92us/step - loss: 0.1491 - accuracy: 0.9570
Epoch 5/6
48000/48000 [==============================] - 4s 91us/step - loss: 0.1370 - accuracy: 0.9597
Epoch 6/6
48000/48000 [==============================] - 4s 90us/step - loss: 0.1308 - accuracy: 0.9616
dense_layer_sizes [64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_19"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_37 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_73 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_38 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_74 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_19 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_37 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_19 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_37 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_75 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_38 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_38 (Dense) (None, 10) 650
________________________________________________________________________________
activation_76 (Activation) (None, 10) 0
================================================================================
Total params: 75,106
Trainable params: 75,106
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 103us/step - loss: 0.4040 - accuracy: 0.8748
Epoch 2/6
48000/48000 [==============================] - 5s 97us/step - loss: 0.2013 - accuracy: 0.9425
Epoch 3/6
48000/48000 [==============================] - 4s 93us/step - loss: 0.1621 - accuracy: 0.9540
Epoch 4/6
48000/48000 [==============================] - 4s 90us/step - loss: 0.1413 - accuracy: 0.9591
Epoch 5/6
48000/48000 [==============================] - 4s 90us/step - loss: 0.1298 - accuracy: 0.9629
Epoch 6/6
48000/48000 [==============================] - 4s 89us/step - loss: 0.1214 - accuracy: 0.9652
dense_layer_sizes [64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_20"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_39 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_77 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_40 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_78 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_20 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_39 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_20 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_39 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_79 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_40 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_40 (Dense) (None, 10) 650
________________________________________________________________________________
activation_80 (Activation) (None, 10) 0
================================================================================
Total params: 75,106
Trainable params: 75,106
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 112us/step - loss: 0.4608 - accuracy: 0.8560
Epoch 2/6
48000/48000 [==============================] - 6s 127us/step - loss: 0.2183 - accuracy: 0.9352
Epoch 3/6
48000/48000 [==============================] - 5s 111us/step - loss: 0.1730 - accuracy: 0.9483
Epoch 4/6
48000/48000 [==============================] - 5s 105us/step - loss: 0.1475 - accuracy: 0.9566
Epoch 5/6
48000/48000 [==============================] - 5s 98us/step - loss: 0.1427 - accuracy: 0.9582
Epoch 6/6
48000/48000 [==============================] - 6s 122us/step - loss: 0.1321 - accuracy: 0.9610
dense_layer_sizes [32, 32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_21"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_41 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_81 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_42 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_82 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_21 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_41 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_21 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_41 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_83 (Activation) (None, 32) 0
________________________________________________________________________________
dense_42 (Dense) (None, 32) 1056
________________________________________________________________________________
activation_84 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_42 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_43 (Dense) (None, 10) 330
________________________________________________________________________________
activation_85 (Activation) (None, 10) 0
================================================================================
Total params: 38,946
Trainable params: 38,946
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 7s 144us/step - loss: 0.5166 - accuracy: 0.8412
Epoch 2/3
48000/48000 [==============================] - 6s 129us/step - loss: 0.2497 - accuracy: 0.9309
Epoch 3/3
48000/48000 [==============================] - 7s 147us/step - loss: 0.2095 - accuracy: 0.9419
dense_layer_sizes [32, 32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_22"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_43 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_86 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_44 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_87 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_22 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_43 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_22 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_44 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_88 (Activation) (None, 32) 0
________________________________________________________________________________
dense_45 (Dense) (None, 32) 1056
________________________________________________________________________________
activation_89 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_44 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_46 (Dense) (None, 10) 330
________________________________________________________________________________
activation_90 (Activation) (None, 10) 0
================================================================================
Total params: 38,946
Trainable params: 38,946
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 6s 135us/step - loss: 0.5476 - accuracy: 0.8271
Epoch 2/3
48000/48000 [==============================] - 8s 168us/step - loss: 0.2757 - accuracy: 0.9202
Epoch 3/3
48000/48000 [==============================] - 5s 102us/step - loss: 0.2231 - accuracy: 0.9364
dense_layer_sizes [32, 32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_23"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_45 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_91 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_46 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_92 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_23 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_45 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_23 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_47 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_93 (Activation) (None, 32) 0
________________________________________________________________________________
dense_48 (Dense) (None, 32) 1056
________________________________________________________________________________
activation_94 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_46 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_49 (Dense) (None, 10) 330
________________________________________________________________________________
activation_95 (Activation) (None, 10) 0
================================================================================
Total params: 38,946
Trainable params: 38,946
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 111us/step - loss: 0.5845 - accuracy: 0.8112
Epoch 2/3
48000/48000 [==============================] - 6s 115us/step - loss: 0.2841 - accuracy: 0.9131
Epoch 3/3
48000/48000 [==============================] - 5s 110us/step - loss: 0.2219 - accuracy: 0.9337
dense_layer_sizes [32, 32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_24"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_47 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_96 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_48 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_97 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_24 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_47 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_24 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_50 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_98 (Activation) (None, 32) 0
________________________________________________________________________________
dense_51 (Dense) (None, 32) 1056
________________________________________________________________________________
activation_99 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_48 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_52 (Dense) (None, 10) 330
________________________________________________________________________________
activation_100 (Activation) (None, 10) 0
================================================================================
Total params: 38,946
Trainable params: 38,946
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 6s 116us/step - loss: 0.5443 - accuracy: 0.8274
Epoch 2/3
48000/48000 [==============================] - 5s 104us/step - loss: 0.2653 - accuracy: 0.9233
Epoch 3/3
48000/48000 [==============================] - 6s 115us/step - loss: 0.2106 - accuracy: 0.9409
dense_layer_sizes [32, 32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_25"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_49 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_101 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_50 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_102 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_25 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_49 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_25 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_53 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_103 (Activation) (None, 32) 0
________________________________________________________________________________
dense_54 (Dense) (None, 32) 1056
________________________________________________________________________________
activation_104 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_50 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_55 (Dense) (None, 10) 330
________________________________________________________________________________
activation_105 (Activation) (None, 10) 0
================================================================================
Total params: 38,946
Trainable params: 38,946
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 114us/step - loss: 0.4992 - accuracy: 0.8474
Epoch 2/3
48000/48000 [==============================] - 5s 111us/step - loss: 0.2467 - accuracy: 0.9303
Epoch 3/3
48000/48000 [==============================] - 5s 104us/step - loss: 0.2019 - accuracy: 0.9448
dense_layer_sizes [32, 32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_26"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_51 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_106 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_52 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_107 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_26 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_51 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_26 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_56 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_108 (Activation) (None, 32) 0
________________________________________________________________________________
dense_57 (Dense) (None, 32) 1056
________________________________________________________________________________
activation_109 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_52 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_58 (Dense) (None, 10) 330
________________________________________________________________________________
activation_110 (Activation) (None, 10) 0
================================================================================
Total params: 38,946
Trainable params: 38,946
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 6s 120us/step - loss: 0.5868 - accuracy: 0.8106
Epoch 2/6
48000/48000 [==============================] - 5s 110us/step - loss: 0.3112 - accuracy: 0.9080
Epoch 3/6
48000/48000 [==============================] - 5s 107us/step - loss: 0.2556 - accuracy: 0.9232
Epoch 4/6
48000/48000 [==============================] - 5s 111us/step - loss: 0.2219 - accuracy: 0.9328
Epoch 5/6
48000/48000 [==============================] - 5s 111us/step - loss: 0.2053 - accuracy: 0.9393
Epoch 6/6
48000/48000 [==============================] - 6s 115us/step - loss: 0.1884 - accuracy: 0.9440
dense_layer_sizes [32, 32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_27"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_53 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_111 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_54 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_112 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_27 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_53 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_27 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_59 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_113 (Activation) (None, 32) 0
________________________________________________________________________________
dense_60 (Dense) (None, 32) 1056
________________________________________________________________________________
activation_114 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_54 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_61 (Dense) (None, 10) 330
________________________________________________________________________________
activation_115 (Activation) (None, 10) 0
================================================================================
Total params: 38,946
Trainable params: 38,946
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 6s 126us/step - loss: 0.5026 - accuracy: 0.8442
Epoch 2/6
48000/48000 [==============================] - 6s 120us/step - loss: 0.2589 - accuracy: 0.9266
Epoch 3/6
48000/48000 [==============================] - 5s 111us/step - loss: 0.2069 - accuracy: 0.9421
Epoch 4/6
48000/48000 [==============================] - 5s 112us/step - loss: 0.1845 - accuracy: 0.9492
Epoch 5/6
48000/48000 [==============================] - 6s 132us/step - loss: 0.1676 - accuracy: 0.9548
Epoch 6/6
48000/48000 [==============================] - 5s 112us/step - loss: 0.1561 - accuracy: 0.9570
dense_layer_sizes [32, 32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_28"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_55 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_116 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_56 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_117 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_28 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_55 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_28 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_62 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_118 (Activation) (None, 32) 0
________________________________________________________________________________
dense_63 (Dense) (None, 32) 1056
________________________________________________________________________________
activation_119 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_56 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_64 (Dense) (None, 10) 330
________________________________________________________________________________
activation_120 (Activation) (None, 10) 0
================================================================================
Total params: 38,946
Trainable params: 38,946
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 7s 150us/step - loss: 0.5850 - accuracy: 0.8099
Epoch 2/6
48000/48000 [==============================] - 5s 112us/step - loss: 0.2859 - accuracy: 0.9157
Epoch 3/6
48000/48000 [==============================] - 7s 142us/step - loss: 0.2297 - accuracy: 0.9351
Epoch 4/6
48000/48000 [==============================] - 7s 154us/step - loss: 0.1945 - accuracy: 0.9457
Epoch 5/6
48000/48000 [==============================] - 7s 144us/step - loss: 0.1696 - accuracy: 0.9522
Epoch 6/6
48000/48000 [==============================] - 5s 111us/step - loss: 0.1580 - accuracy: 0.9556
dense_layer_sizes [32, 32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_29"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_57 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_121 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_58 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_122 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_29 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_57 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_29 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_65 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_123 (Activation) (None, 32) 0
________________________________________________________________________________
dense_66 (Dense) (None, 32) 1056
________________________________________________________________________________
activation_124 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_58 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_67 (Dense) (None, 10) 330
________________________________________________________________________________
activation_125 (Activation) (None, 10) 0
==========================================================#61;======================
Total params: 38,946
Trainable params: 38,946
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 6s 121us/step - loss: 0.5203 - accuracy: 0.8362
Epoch 2/6
48000/48000 [==============================] - 5s 110us/step - loss: 0.2587 - accuracy: 0.9243
Epoch 3/6
48000/48000 [==============================] - 5s 106us/step - loss: 0.2045 - accuracy: 0.9410
Epoch 4/6
48000/48000 [==============================] - 5s 104us/step - loss: 0.1789 - accuracy: 0.9501
Epoch 5/6
48000/48000 [==============================] - 5s 108us/step - loss: 0.1637 - accuracy: 0.9555
Epoch 6/6
48000/48000 [==============================] - 5s 106us/step - loss: 0.1535 - accuracy: 0.9581
dense_layer_sizes [32, 32]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_30"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_59 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_126 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_60 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_127 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_30 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_59 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_30 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_68 (Dense) (None, 32) 36896
________________________________________________________________________________
activation_128 (Activation) (None, 32) 0
________________________________________________________________________________
dense_69 (Dense) (None, 32) 1056
________________________________________________________________________________
activation_129 (Activation) (None, 32) 0
________________________________________________________________________________
dropout_60 (Dropout) (None, 32) 0
________________________________________________________________________________
dense_70 (Dense) (None, 10) 330
________________________________________________________________________________
activation_130 (Activation) (None, 10) 0
================================================================================
Total params: 38,946
Trainable params: 38,946
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 108us/step - loss: 0.6101 - accuracy: 0.8081
Epoch 2/6
48000/48000 [==============================] - 5s 110us/step - loss: 0.2820 - accuracy: 0.9218
Epoch 3/6
48000/48000 [==============================] - 6s 124us/step - loss: 0.2187 - accuracy: 0.9398
Epoch 4/6
48000/48000 [==============================] - 5s 108us/step - loss: 0.1816 - accuracy: 0.9501
Epoch 5/6
48000/48000 [==============================] - 5s 114us/step - loss: 0.1606 - accuracy: 0.9565
Epoch 6/6
48000/48000 [==============================] - 5s 107us/step - loss: 0.1487 - accuracy: 0.9589
dense_layer_sizes [64, 64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_31"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_61 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_131 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_62 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_132 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_31 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_61 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_31 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_71 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_133 (Activation) (None, 64) 0
________________________________________________________________________________
dense_72 (Dense) (None, 64) 4160
________________________________________________________________________________
activation_134 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_62 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_73 (Dense) (None, 10) 650
________________________________________________________________________________
activation_135 (Activation) (None, 10) 0
================================================================================
Total params: 79,266
Trainable params: 79,266
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 6s 115us/step - loss: 0.3673 - accuracy: 0.8900
Epoch 2/3
48000/48000 [==============================] - 5s 104us/step - loss: 0.1600 - accuracy: 0.9555
Epoch 3/3
48000/48000 [==============================] - 5s 107us/step - loss: 0.1213 - accuracy: 0.9665
dense_layer_sizes [64, 64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_32"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_63 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_136 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_64 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_137 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_32 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_63 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_32 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_74 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_138 (Activation) (None, 64) 0
________________________________________________________________________________
dense_75 (Dense) (None, 64) 4160
________________________________________________________________________________
activation_139 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_64 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_76 (Dense) (None, 10) 650
________________________________________________________________________________
activation_140 (Activation) (None, 10) 0
================================================================================
Total params: 79,266
Trainable params: 79,266
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 110us/step - loss: 0.3320 - accuracy: 0.8991
Epoch 2/3
48000/48000 [==============================] - 5s 108us/step - loss: 0.1442 - accuracy: 0.9607
Epoch 3/3
48000/48000 [==============================] - 5s 103us/step - loss: 0.1134 - accuracy: 0.9695
dense_layer_sizes [64, 64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_33"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_65 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_141 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_66 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_142 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_33 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_65 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_33 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_77 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_143 (Activation) (None, 64) 0
________________________________________________________________________________
dense_78 (Dense) (None, 64) 4160
________________________________________________________________________________
activation_144 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_66 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_79 (Dense) (None, 10) 650
________________________________________________________________________________
activation_145 (Activation) (None, 10) 0
================================================================================
Total params: 79,266
Trainable params: 79,266
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 6s 118us/step - loss: 0.3735 - accuracy: 0.8865
Epoch 2/3
48000/48000 [==============================] - 5s 109us/step - loss: 0.1466 - accuracy: 0.9615
Epoch 3/3
48000/48000 [==============================] - 6s 117us/step - loss: 0.1154 - accuracy: 0.9688
dense_layer_sizes [64, 64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_34"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_67 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_146 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_68 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_147 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_34 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_67 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_34 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_80 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_148 (Activation) (None, 64) 0
________________________________________________________________________________
dense_81 (Dense) (None, 64) 4160
________________________________________________________________________________
activation_149 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_68 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_82 (Dense) (None, 10) 650
________________________________________________________________________________
activation_150 (Activation) (None, 10) 0
================================================================================
Total params: 79,266
Trainable params: 79,266
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 105us/step - loss: 0.3600 - accuracy: 0.8920
Epoch 2/3
48000/48000 [==============================] - 5s 100us/step - loss: 0.1543 - accuracy: 0.9586
Epoch 3/3
48000/48000 [==============================] - 5s 106us/step - loss: 0.1130 - accuracy: 0.9692
dense_layer_sizes [64, 64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_35"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_69 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_151 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_70 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_152 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_35 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_69 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_35 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_83 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_153 (Activation) (None, 64) 0
________________________________________________________________________________
dense_84 (Dense) (None, 64) 4160
________________________________________________________________________________
activation_154 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_70 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_85 (Dense) (None, 10) 650
________________________________________________________________________________
activation_155 (Activation) (None, 10) 0
================================================================================
Total params: 79,266
Trainable params: 79,266
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/3
48000/48000 [==============================] - 5s 113us/step - loss: 0.3310 - accuracy: 0.9006
Epoch 2/3
48000/48000 [==============================] - 5s 104us/step - loss: 0.1388 - accuracy: 0.9621
Epoch 3/3
48000/48000 [==============================] - 5s 106us/step - loss: 0.1067 - accuracy: 0.9707
dense_layer_sizes [64, 64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_36"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_71 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_156 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_72 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_157 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_36 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_71 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_36 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_86 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_158 (Activation) (None, 64) 0
________________________________________________________________________________
dense_87 (Dense) (None, 64) 4160
________________________________________________________________________________
activation_159 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_72 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_88 (Dense) (None, 10) 650
________________________________________________________________________________
activation_160 (Activation) (None, 10) 0
================================================================================
Total params: 79,266
Trainable params: 79,266
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 108us/step - loss: 0.3465 - accuracy: 0.8963
Epoch 2/6
48000/48000 [==============================] - 5s 105us/step - loss: 0.1464 - accuracy: 0.9599
Epoch 3/6
48000/48000 [==============================] - 5s 102us/step - loss: 0.1106 - accuracy: 0.9700
Epoch 4/6
48000/48000 [==============================] - 5s 102us/step - loss: 0.0940 - accuracy: 0.9744
Epoch 5/6
48000/48000 [==============================] - 5s 101us/step - loss: 0.0843 - accuracy: 0.9767
Epoch 6/6
48000/48000 [==============================] - 5s 103us/step - loss: 0.0760 - accuracy: 0.9802
dense_layer_sizes [64, 64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_37"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_73 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_161 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_74 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_162 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_37 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_73 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_37 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_89 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_163 (Activation) (None, 64) 0
________________________________________________________________________________
dense_90 (Dense) (None, 64) 4160
________________________________________________________________________________
activation_164 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_74 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_91 (Dense) (None, 10) 650
________________________________________________________________________________
activation_165 (Activation) (None, 10) 0
================================================================================
Total params: 79,266
Trainable params: 79,266
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 110us/step - loss: 0.3692 - accuracy: 0.8888
Epoch 2/6
48000/48000 [==============================] - 5s 104us/step - loss: 0.1547 - accuracy: 0.9573
Epoch 3/6
48000/48000 [==============================] - 5s 103us/step - loss: 0.1164 - accuracy: 0.9682
Epoch 4/6
48000/48000 [==============================] - 5s 106us/step - loss: 0.0960 - accuracy: 0.9739
Epoch 5/6
48000/48000 [==============================] - 5s 105us/step - loss: 0.0855 - accuracy: 0.9771
Epoch 6/6
48000/48000 [==============================] - 5s 105us/step - loss: 0.0799 - accuracy: 0.9787
dense_layer_sizes [64, 64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_38"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_75 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_166 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_76 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_167 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_38 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_75 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_38 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_92 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_168 (Activation) (None, 64) 0
________________________________________________________________________________
dense_93 (Dense) (None, 64) 4160
________________________________________________________________________________
activation_169 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_76 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_94 (Dense) (None, 10) 650
________________________________________________________________________________
activation_170 (Activation) (None, 10) 0
================================================================================
Total params: 79,266
Trainable params: 79,266
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 109us/step - loss: 0.3541 - accuracy: 0.8931
Epoch 2/6
48000/48000 [==============================] - 5s 104us/step - loss: 0.1614 - accuracy: 0.9559
Epoch 3/6
48000/48000 [==============================] - 5s 103us/step - loss: 0.1266 - accuracy: 0.9664
Epoch 4/6
48000/48000 [==============================] - 5s 102us/step - loss: 0.0996 - accuracy: 0.9723
Epoch 5/6
48000/48000 [==============================] - 5s 106us/step - loss: 0.0907 - accuracy: 0.9759
Epoch 6/6
48000/48000 [==============================] - 5s 101us/step - loss: 0.0828 - accuracy: 0.9781
dense_layer_sizes [64, 64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_39"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_77 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_171 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_78 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_172 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_39 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_77 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_39 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_95 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_173 (Activation) (None, 64) 0
________________________________________________________________________________
dense_96 (Dense) (None, 64) 4160
________________________________________________________________________________
activation_174 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_78 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_97 (Dense) (None, 10) 650
________________________________________________________________________________
activation_175 (Activation) (None, 10) 0
================================================================================
Total params: 79,266
Trainable params: 79,266
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 7s 138us/step - loss: 0.3629 - accuracy: 0.8934
Epoch 2/6
48000/48000 [==============================] - 6s 128us/step - loss: 0.1662 - accuracy: 0.9545
Epoch 3/6
48000/48000 [==============================] - 5s 102us/step - loss: 0.1245 - accuracy: 0.9667
Epoch 4/6
48000/48000 [==============================] - 5s 99us/step - loss: 0.1035 - accuracy: 0.9722
Epoch 5/6
48000/48000 [==============================] - 5s 105us/step - loss: 0.0910 - accuracy: 0.9767
Epoch 6/6
48000/48000 [==============================] - 5s 100us/step - loss: 0.0820 - accuracy: 0.9787
dense_layer_sizes [64, 64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_40"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_79 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_176 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_80 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_177 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_40 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_79 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_40 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_98 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_178 (Activation) (None, 64) 0
________________________________________________________________________________
dense_99 (Dense) (None, 64) 4160
________________________________________________________________________________
activation_179 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_80 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_100 (Dense) (None, 10) 650
________________________________________________________________________________
activation_180 (Activation) (None, 10) 0
================================================================================
Total params: 79,266
Trainable params: 79,266
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
48000/48000 [==============================] - 5s 110us/step - loss: 0.3803 - accuracy: 0.8851
Epoch 2/6
48000/48000 [==============================] - 5s 102us/step - loss: 0.1589 - accuracy: 0.9570
Epoch 3/6
48000/48000 [==============================] - 6s 125us/step - loss: 0.1235 - accuracy: 0.9664
Epoch 4/6
48000/48000 [==============================] - 5s 102us/step - loss: 0.1064 - accuracy: 0.9710
Epoch 5/6
48000/48000 [==============================] - 5s 103us/step - loss: 0.0934 - accuracy: 0.9753
Epoch 6/6
48000/48000 [==============================] - 5s 103us/step - loss: 0.0875 - accuracy: 0.9760
dense_layer_sizes [64, 64]
filters 8
kernel_size 3
pool_size 2
Model: "sequential_41"
________________________________________________________________________________
Layer (type) Output Shape Param #
================================================================================
conv2d_81 (Conv2D) (None, 26, 26, 8) 80
________________________________________________________________________________
activation_181 (Activation) (None, 26, 26, 8) 0
________________________________________________________________________________
conv2d_82 (Conv2D) (None, 24, 24, 8) 584
________________________________________________________________________________
activation_182 (Activation) (None, 24, 24, 8) 0
________________________________________________________________________________
max_pooling2d_41 (MaxPooling2D) (None, 12, 12, 8) 0
________________________________________________________________________________
dropout_81 (Dropout) (None, 12, 12, 8) 0
________________________________________________________________________________
flatten_41 (Flatten) (None, 1152) 0
________________________________________________________________________________
dense_101 (Dense) (None, 64) 73792
________________________________________________________________________________
activation_183 (Activation) (None, 64) 0
________________________________________________________________________________
dense_102 (Dense) (None, 64) 4160
________________________________________________________________________________
activation_184 (Activation) (None, 64) 0
________________________________________________________________________________
dropout_82 (Dropout) (None, 64) 0
________________________________________________________________________________
dense_103 (Dense) (None, 10) 650
________________________________________________________________________________
activation_185 (Activation) (None, 10) 0
================================================================================
Total params: 79,266
Trainable params: 79,266
Non-trainable params: 0
________________________________________________________________________________
Epoch 1/6
60000/60000 [==============================] - 8s 129us/step - loss: 0.3397 - accuracy: 0.8988
Epoch 2/6
60000/60000 [==============================] - 7s 124us/step - loss: 0.1434 - accuracy: 0.9614
Epoch 3/6
60000/60000 [==============================] - 6s 100us/step - loss: 0.1131 - accuracy: 0.9692
Epoch 4/6
60000/60000 [==============================] - 6s 107us/step - loss: 0.0947 - accuracy: 0.9740
Epoch 5/6
60000/60000 [==============================] - 6s 102us/step - loss: 0.0839 - accuracy: 0.9781
Epoch 6/6
60000/60000 [==============================] - 6s 102us/step - loss: 0.0767 - accuracy: 0.9796
The parameters of the best model are:
{'dense_layer_sizes': [64, 64], 'epochs': 6, 'filters': 8, 'kernel_size': 3, 'pool_size': 2}
10000/10000 [==============================] - 1s 73us/step
loss : 0.047322944842268046
accuracy : 0.9843999743461609