CS229 LECTURE 1 -- 机器学习的动机及运用

This course is orgnized into four major sections:

Supervised learning

We are giving the algorithm a bunch of “right answers”, and we are expecting the algorithm to provide us with right answers using the existing data.(e.g. regression 回归问题;classification 分类问题)

Regression: 房价预测

Regression: e.g. 房价预测

Classification: 多数情况下模型是离散的，比如根据肿瘤大小预测肿瘤是否为良性（0 vs. 1）

Classification: 多数情况下模型是离散的，比如根据肿瘤大小预测肿瘤是否为良性（0 vs. 1）

Learning Theory

How and why these learning models work.

It helps us better understand and better use machine learning.

Unsupervised Learning

You need to figure out what the structure is in a given data set when you are not given the right answers.(e.g. clustering 聚类分析)

The “tumor” example of unsupervised learning

聚类分析在图像处理方面的应用（聚类分析常常被用于图像处理）

Cocktail party problem

Reinforcement Learning

You are asked to make a sequence of decisions over time.

Reward function: “bad” dog and “good” dog!

Applied to many problems in robotics, web crawling and so on.

Applications of reinforcement learning