MIT ML 听课笔记（一）

machine Learning : 

• supervised learning ; 
• unsupervised Learning

supervised learning:对于数据集中的每个数据，都有相应的正确答案。算法就是基于这些来做出预测。

  1.regression problem (predict continuous valued output):

Regression is a ML task where T has a real-valued outcome on some continuous sub-space 

  2.classification problem （预测是否为某种离散值）

Classification is a ML task where T has a discrete set of outcomes

    

m=number of training examples

x’s= input variable/features

y’s=output variable/target variable

     (x,y) one training example

:=means assignment.  

a:=b means take the value in b and use it overwrite whatever value is a . This means set a to be equal to the value of b, which is assignment.

运用梯度下降的两个技巧：

1. feature scaling :特征缩放（讲特征范围控制在-3—3）

mean normalization： replace xi with xi-ui to make features have approximately zero mean

x <-  x-u/s

u: average value of x in training set

s: range(max-min)

不需要太精确，只是为了让梯度下降法更快

2.

如何确保梯度下降算法work correctly：

for sufficiently small alpha, j-theta should decrease on every iteration. （cost一直在减小）

所以如果j(theta)变大或者没有一直变小，往往是learning rate太大了，需要变小。

如果alpha过于小，那就会收敛的很慢。