Deep neural networks form an important
subfield of machine learning that is responsible for much of
the progress in in cognitive computing in recent years in
areas of computer vision, audio processing, and natural
language processing. Deep networks can be trained with a
single endtoend model and bypass the need for
traditional taskspecific feature engineering. In this way
deep learning simplifies learning tasks and allows using
developed models to new tasks. Deep networks are suitable for
parallel processing implementations and can easily leverage
intensive computational resources. The course will focus on
mathematical concepts, numerical algorithms, principles, GPU
frameworks, and applications of deep learning. Topics include
deep feedforward networks, convolutional networks,
sequence modeling, deep generative models, and deep
reinforcement learning with applications to data analysis,
computer vision, and natural language processing. Several
programming assignments and a project will practice the
application of deep learning techniques to actual problems. A
dedicated cluster will be used to support course assignments.
The course requires sufficient math and programming background
but does not require prior knowledge in machine learning. For
further details please refer to the course website or contact
the course instructor.
CS57701: (SB104)
CS57702: (Internet)
Class hours:
Tuesday, Thursday
5:006:15pm
Deep learning can be covered at different levels. The focus
of this course is principles, mathematical concepts,
algorithms, and techniques used in deep learning. Students
in the course are expected to write computer programs
implementing different techniques taught in the course. The
course requires mathematical background and some programming
experience. This course does not
intend to teach how to use a specific application
software but will use a GPU frameworks for the assignments.
1. Introduction to machine learning and computational foundations
component

description

weight

participation

up to 4 unjustified missed
classes and all quizes ⇒
full credit

5%

assignments

45 TBD

25%

project

presentation (5%) project
(15%)

20%

midterm exam

open notes (1 double sided
8.5x11" page)

10%

final exam

open notes (2 double sided
8.5x11" pages)

40%

total


100%

class  date  topic 
assignment 
1  01/13  Introduction  AS0 
2  01/15  
3  01/20  No class (MLK day)  
4  01/22  Introduction to GPU frameworks  
5  01/27  AS1  
6  01/29  Neural networks  
7  02/03  
8  02/05  
9  02/10  Deep feedforward networks  
10  02/12  AS2  
11  02/17  
12  02/19  Regularization and optimization  
13  02/24  
14  02/26  
15  03/02  Convolutional networks  PROJ 
16  03/04  Midterm  
17  03/09  
18  03/11  AS3  
19  03/16  Spring break  
20  03/18  Spring break  
21  03/23  Representation learning  
22  03/25  AS4  
23  03/30  Recurrent networks  
24  04/01  
25  04/06  Generative models  
26  04/08  AS5  
27  04/13  Deep reinforcement learning  
28  04/15  
29  04/20  Presentations  
30  04/22  
31  04/27  
32  04/29  
33  05/04  Final exam 57pm 
Videos of lectures are available through blackboard
Topic 
Reading 
Introduction 
Ch. 13, 5 
Neural networks 
Ch. 4 
Deep neural networks 
Ch. 6 
Regularization and optimization 
Ch. 78 
Convolutional networks 
Ch. 9 
Representation learning 
Ch. 1415 
Recurrent networks 
Ch. 10 
Generative models 
Ch. 20 
Deep reinforcement learning 
Assignment  Description  Data  Weight  Due date 
assignment 0 
basic review 
none 
0% 

assignment 1  5%  
assignment 2  5%  
assignment 3 
5%  
assignment 4 
5%  
assignment 5 
5%  
project  presentation
(10%)
project (10%) 
N/A  20%  (proposal ) (final submission) 
Additional assignments: