Machine Learning - Lecture 16

Reinforcement Learning (R.L.)

① MDPs (Markov Decision Processes)

② Value Functions

③ Value Iteration

④ Policy Iteration

(both ③ and ④ are algorithms for solving R.L. problems)

Supervised Learning: we have the training set in which we were given sort of the right answer of every training example and it was the just a drop of the learning algorithms to replicate more of the right answers.

Unsupervised Learning: we had just a bunch of unlabeled data just the x's and it was the job in the learning alogrithm to discover so-called structure in the data and several algorithms like cluster analysis K-means, a mixture of all the sort PCA, ICA and so on.

Today we just talk about a different class of learning algorithms between supervised and unsupervised — R.L.

there's a helicopter experiment performed by Andrew Ng at Stanford University(you could see the video and the details of that experiment on the Internet), which is a unmanned helicopter controlld by R.L. algorithms.

It's different from Supervised Learning, because usually we actually do not konw