课程主页: https://cs229.stanford.edu/
课程资料: https://cs229.stanford.edu/syllabus-autumn2018.html
课程资料备份: https://github.com/maxim5/cs229-2018-autumn
课程回放1（配合双字幕插件食用）: https://youtu.be/jGwO_UgTS7I
课程回放2（无追踪短链接）: https://b23.wtf/gRIqDIx
学习笔记Offical: https://stanford.edu/~shervine/teaching/cs-229/
学习笔记Offical（简中）: https://stanford.edu/~shervine/l/zh/teaching/cs-229/
（开新坑？希望三个月能弄完然后学CS230😢）
Lecture 1 - Logistics and Introduction
Machine Learning Define
Machine Learning(Arthur Samuel, 1959): Field of study that gives computersthe ability to learn without being explicitly programmed.
Well-posed Learning Problem(Tom Mitchell, 1998): A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P,improves with experience E.
Supervised Learning
You are given a datasets with inputs X and lables Y and your goal is to learn a mapping from X to Y.
regression
the term regression refers to that the value Y you’re trying to predict is continuous.
classfication
the term classification refers to the Y takes on a discrete number of variables. (the output is discrete)
(the top symbol is a different way of visualizing the same data)
classfication - muliple features
two features
more features in reality problem
[VEDIO]
X: road state / Driver direction
Y: steering direction
X: road state / Arbitrator Direction
Y1: steering direction in one -lane road
Y2：steering direction in two-lane road
Machine Learning Strategy(Learning Theory)
Learn more systematic engineering principles for machine learning to efficiently figure out what to do next.
Deep learning
It’s one subset of machine learning that’s really hot right now.
CS330 more narrowly covers just deep learning.
Unsupervised learning
You given a dataset with no labels, just given inputs X and no Y and your goal is find something interesting(eg. figure out the structure in this data) about that.
some unsupervised learning problem example:
news clusters / genetic microarray data / organize computing clusters / social network cohesive communities analysis / market segmentation / astronomical data analysis: clustering problem
cocktail party problem: a ICA(Independent Componenets Analysis) problem
learning analogies: man is to woman as king is to queen, Tokyo is to Japan as Washington DC is to the United States. You can learning this relation just from texts on the Internet.
Reinforcement learning
Giving signals called a reward signal, you can have a learning algorithm figure out by itself how to optimize the reward.
For example: you can’t teach a dog the optimal way to behave(you aren’t a dog), but you can go “Oh, good dog” whenever it behavs well, go “bad dog” when it misbeaves. Over time, the dog learns to do more of the good dog things and fewer of the bad dog things.