Department of Computer Science, National Chiao-Tung University

IOC5184(5259) Deep Learning and Practice

Ÿ   Time of Offering: Spring Term, 2017

Ÿ   Level: Graduate Students

Ÿ   Prerequisites: The recommended prerequisites are to have taken Linear Algebra and Probability Theory and Machine Learning (suggested).

Ÿ   Course Instructor

­         Wen-Hsiao Peng (彭文孝), Ph.D.

­         E-mail: wpeng@cs.nctu.edu.tw

­         Office: EC431 (工三館431)

­         Phone: Ext56625

­         Lab: Multimedia Architecture and Processing Laboratory (MAPL)

­         URL: http://mapl.nctu.edu.tw

Ÿ   Teaching Assistant

­         Wei-Zhi Pan (潘韋志) will81917@gmail.com

­         Huang Jinbo (黃勁博) iceman1216i@gmail.com

­         Zhang Jiaren (張家仁) followwar@gmail.com

­         Room: EC 621 (工程三館)

­         Phone: 56639

Ÿ   Course Homepage

­         http://mapl.nctu.edu.tw/course/DL_2017/index.php

Ÿ   Lectures

­         The course meets on Tuesdays from 18:30 to 21:20 (2IJK) in EC220 and Thursdays from 12:20 to 15:10 (4XEF) in EC022.

Ÿ   Course Outline

1.          Introduction

·      Moments & Moments Generating Functions

·      Chebyshev and Schwarz Inequality

·      Chernoff Bound

·      Characteristic Functions

·      Estimator for Mean and Variance of the Normal Law

2.          Machine Learning Basics

·      Linear Algebra

·      Probability and Information Theory

·      Numerical Computation

·      Machine Learning Basics

3.          Deep Networks

·      Deep Feedforward Networks

·      Convolutional Networks

·      Optimization for Training Deep Models

·      Regularization for Deep Learning

·      Recurrent and Recursive Nets

4.          Deep Learning Research

·      Linear Factor Models

·      Autoencoders

·      Representation Learning

·      Structured Probabilistic Models for Deep Learning

·      Approximate Inference

·      Deep Generative Models

5.          Deep Reinforcement Learning

·      Intro. to RL

·      MDP/POMDP + TD Learning

·      Policy Gradient + DQN

·      DQN Applications: Atari, AlphaGo and Robotics.

6.          Transfer Learning and Applications

7.          Paper Study and Presentation

Ÿ   Reference

­         I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning, 1st Ed., MIT Press, Dec. 2016

Ÿ   Grading Policy

­         Computer assignments 40%

­         Paper presentation 20%

­         Final project and presentation 20%

­         Final written exam 20%