Wei Lin @ PKU

00103335: Deep Learning and Reinforcement Learning

Course Description

As highly successful and widely applied machine learning methods, deep learning and reinforcement learning are the core techniques underlying the latest major breakthroughs in the field of AI. Building on the general principles and methodology of machine learning and motivated by important practical problems, this course will introduce the basic concepts and methods, mathematical foundations and theory, optimization algorithms, and applications and case studies of deep learning and reinforcement learning. The part on deep learning will cover feedforward neural networks, regularization and optimization for deep learning, convolutional neural networks, recurrent neural networks, and autoencoders and generative models; the part on reinforcement learning will cover multi-armed bandits, Markov decision processes, dynamic programming, Monte Carlo methods, temporal difference learning, and deep reinforcement learning.

Syllabus

Lectures and Assignments

Week Date Topics References Assignments Notes
1 9/8 Introduction FML Chap. 1, UML Chap. 1, DL Sec. 5.1
9/10 No-free-lunch theorem, bias–variance trade-off UML Chap. 5, DL Secs. 5.2–5.4
2 9/17 PAC framework, finite hypthesis sets FML Chap. 2 FML 2.1, 2.3, 2.7, 2.9, 2.10, 2.12
3 9/22 Rademacher complexity FML Sec. 3.1
9/24 Growth function, VC dimension FML Secs. 3.2, 3.3 FML 3.2, 3.4, 3.8, 3.12, 3.16, 3.17, 3.23, 3.24, 3.31; supplementary problems Homework 1 due 10/15
4 10/1 No class
5 10/6 No class
10/8 No class
6 10/15 Lower bounds, feedforward networks FML Sec. 3.4, DL Secs. 5.11, 6.1–6.3
7 10/20 Approximation theory DL Sec. 6.4, UML Secs. 20.3, 20.4, Leshno et al. (1993)
10/22 Backprop, explicit regularization DL Secs. 6.5, 7.1–7.3
8 10/29 Implicit regularization DL Secs. 7.4–7.14
9 11/3 Optimization for DL DL Chap. 8
11/5
10 11/12
11 11/17
11/19
12 11/26 Midterm exam
13 12/1
12/3
14 12/10
15 12/15
12/17 Oral presentations
16 12/24 Oral presentations