来源于coursea 的公开课
A*B 一般意义的矩阵相乘
A.*B矩阵各位相乘
A.^2 A矩阵的每个数平方
1./A 对A矩阵的各位取倒
.表示对每一项都如此操作
log （A）
ｅｘp（Ａ）
abs(A)
-A th
v+ones(length(v),1) add vector of all ones to v
or v+1
A' is a transpose
max(A) max vluae of a
[max,ind]=max(a) the first are the value,and the second is index
a<3 comparsion for each element in matrix
fina(a<3) tell me which element is less than 3
magic function return a magic matrix
sum(A)
prod meand product (乘积 )
ceil(A)
floor()
max(rand(3）,rand())
max(a,[],1) max of each colom
max(a,[],2) max of each row
A(:) every element in a vector
sum(A,1) sum of 1 colum of Ai
sum(A,2) sum of each row of A
flipud(A) A的山下颠倒
plot(x,y) 以第一个为横轴，第二个为纵轴绘图
hold on
plot(x,y,'r') 追加内容
xlebel(""）
ylebel("")
title("")
print -dpng "my.png" save as a file in png format
figure(1);plot (t,y1) 分割子图
axis([0.5 1 -1 1]) x range form 0.5 to 1 t range from -1 to 1
clf clear the figure
imagicesc(A) 生成彩色块图
colorbar 彩色边栏对应图
colormap gray 灰白度图 使用 colormap("list“） look for what's avaliable
zero(n,m) return a matrix that nXm elemetary are the 0
for i=1:10,
code
end;
indecse=1:10
for i=indecs the same as above
exit
quit
function y=mfunc(x)
y=x^2
addpath('d:ml')
function [y1,y2]=mcun1(x)
return multipule values
好像必须用单引号
theta regularition 时候不考虑theta0 ;
nn grad 的时候 g(Z1) 我第一次将 A2带入g得不到正确的结果
When training neural networks, it is important to randomly initialize the parameters for symmetry breaking
2One effective strategy for choosing init is to base it on the number of units in the network. A good choice of init is init = √6/(√Lin+Lout), where Lin = sl and Lout = sl+1 are the number of units in the layers adjacent to Θ(l
mu = mean(X);
X_norm = bsxfun(@minus, X, mu);
sigma = std(X_norm);
X_norm = bsxfun(@rdivide, X_norm, sigma);
bsxfun 函数的作用
plot 函数只在vector相等时画图 不相等会报错，也不绘图。