Fall 2022
3 Credit
Hours
Description:
As machine learning (ML) and Artificial Intelligence (AI) gets rapidly adopted
everywhere, the speed in learning and inference has become one of the most
frequently encountered roadblocks for practical adoptions. This course focuses
on the challenges and solutions for achieving high performance and real-time
response of Deep Learning-based ML and AI while keeping the accuracy
satisfactory. The course will cover the common platforms and toolkits for
modern Machine Learning and AI, the factors and tradeoffs affecting the
performance, various optimization techniques for ML and AI, and the trends and
research directions being actively investigated in this field.
By the
end of the course, students should be able to do the following.
·
explain the main techniques for optimizing
the speed of AI and Deep Learning
·
explain the factors and tradeoffs affecting
the performance
·
master some existing tools and techniques for
addressing the issues
·
describe the directions being actively
investigated in this field
The course will
consist of a series of lectures, guest lectures, student presentations, and
in-class discussions.
Xipeng Shen
(xshen5) - Instructor
Email: xshen5@ncsu.edu
Phone: 919-513-7577
Office Location: EBII 3276
Days:
Friday
Time: 8:30am - 11:15am
Campus: Centennial
Location: EBIII 02201
This meeting is required.
None. The class will be based on slides and research papers.
None.
None.
Introduction-level knowledge of ML, data
structures (CSC316 or equivalent), Python. Helpful: C/C++, CUDA, GPU.
None.
None.
This course does
not fulfill a General Education Program category.
This course does
not fulfill a General Education Program co-requisite.
This course will
not require students to provide their own transportation. Non-scheduled class
time for field trips or out-of-class activities is NOT required for this class.
None.
Component |
Weight |
Details |
Course Project |
25% |
The course project will be a group programming
project with group size as two. |
Unit Projects |
50% |
There will be five small programming
assignments. |
Presentation |
15% |
A 2 or 3-student group will give one
presentation on a research paper. CSC791 students are required to write a
review of the paper. |
Class participation |
10% |
Students are expected to prepare and actively
participate class discussions. |
Optional project |
10% |
Students have the
opportunity to conduct an optional programming assignment. |
This Course uses the
Following (Non-Standard) Letter Grading Scale:
97% |
≤ |
A+ |
≤ |
100% |
93% |
≤ |
A |
< |
97% |
87% |
≤ |
A- |
< |
93% |
85% |
≤ |
B+ |
< |
87% |
80% |
≤ |
B |
< |
85% |
75% |
≤ |
B- |
< |
80% |
70% |
≤ |
C+ |
< |
75% |
65% |
≤ |
C |
< |
70% |
60% |
≤ |
C- |
< |
65% |
56% |
≤ |
D+ |
< |
60% |
50% |
≤ |
D |
< |
56% |
45% |
≤ |
D- |
< |
50% |
0% |
≤ |
F |
< |
45% |
Performance in
research, seminar and independent study types of
courses (6xx and 8xx) is evaluated as either "S" (Satisfactory) or
"U" (Unsatisfactory), and these grades are not used in computing the
grade point average. For credit only courses (S/U) the requirements necessary
to obtain the grade of "S" must be clearly outlined.
Information about
and requirements for auditing a course can be found at http://policies.ncsu.edu/regulation/reg-02-20-04.
If an extended
deadline is not authorized by the Graduate School, an unfinished incomplete
grade will automatically change to an F after either (a) the end of the next
regular semester in which the student is enrolled (not including summer
sessions), or (b) by the end of 12 months if the student is not enrolled,
whichever is shorter. Incompletes that change to F will count as an attempted
course on transcripts. The burden of fulfilling an incomplete grade is the
responsibility of the student. The university policy on incomplete grades is
located at http://policies.ncsu.edu/regulation/reg-02-50-03.
Additional information relative to incomplete grades for graduate students can
be found in the Graduate Administrative Handbook in Section 3.18.F at http://www.fis.ncsu.edu/grad_publicns/handbook/
No late assignments will be accepted. If an emergency (e.g.,
hospitalization) prevented an assignment from being submitted, its grade will
be determined by averaging the completed assignments in the same category.
Instructor may grant an extension if a student has some extenuating
circumstances warranting it (which is very rare). Student must receive
this extension by noon the day it is due. Note: requesting an extension is not
the same as receiving it.
For complete attendance and excused absence
policies, please see http://policies.ncsu.edu/regulation/reg-02-20-03
While attendance is not taken, students are
expected to attend. Class periods will contain material that is not in the
reading assignments that students are responsible for knowing. While every effort
is made to provide critical information via electronic resources (web page,
mailing list, etc.), some information may not show up outside class
in a timely manner (or at all). Although this is accidental, students are
nevertheless responsible for all information presented in class.
None.
None.
None.
Students are
required to comply with the university policy on academic integrity found in
the Code of Student Conduct found at http://policies.ncsu.edu/policy/pol-11-35-01
Students are
expected to maintain high standards of academic integrity and
honesty. University guidelines regarding academic integrity will be
followed. Cheating will result in disciplinary actions, up to the full
penalties specified in the guideline. A grade of zero will be given to the
assignment in question for a minor offense. A major offense, including any
violation on a test, could result in failure of the course. All suspected
violations will be reported to the Office of Student Conduct,
where a guilty outcome on a second offense can mean suspension from the
university.
Please keep in mind that academic integrity
in the classroom translates to professional integrity in the workplace.
Moreover, awarding similar grades to students who have maintained academic
integrity and to students who have cheated results in conferring equivalent
degrees on them, and reduces the value of that degree in the workplace. It is
the responsibility of every student as well as the instructor and TA to see
that this is not allowed to happen.
See http://policies.ncsu.edu/policy/pol-11-35-01
for a detailed explanation of academic honesty.
None.
Your signature on
any test or assignment indicates "I have neither given nor received
unauthorized aid on this test or assignment."
Students may be
required to disclose personally identifiable information to other students in
the course, via electronic tools like email or web-postings, where relevant to
the course. Examples include online discussions of class topics,
and posting of student coursework. All students are expected to respect
the privacy of each other by not sharing or using such information outside the
course.
Electronically-hosted Components: online discussions of class topics
Reasonable
accommodations will be made for students with verifiable disabilities. In order to take advantage of available accommodations,
student must register with the Disability Services Office (http://www.ncsu.edu/dso), 919-515-7653. For
more information on NC State's policy on working with students with
disabilities, please see the Academic Accommodations for Students with
Disabilities Regulation at http://policies.ncsu.edu/regulation/reg-02-20-01.
NC State
University provides equality of opportunity in education and employment for all
students and employees. Accordingly, NC State affirms its commitment to
maintain a work environment for all employees and an academic environment for
all students that is free from all forms of discrimination. Discrimination
based on race, color, religion, creed, sex, national origin, age, disability,
veteran status, or sexual orientation is a violation of state and federal law
and/or NC State University policy and will not be tolerated. Harassment of any
person (either in the form of quid pro quo or creation of a hostile
environment) based on race, color, religion, creed, sex, national origin, age,
disability, veteran status, or sexual orientation also is a violation of state
and federal law and/or NC State University policy and will not be tolerated.
Retaliation against any person who complains about discrimination is also
prohibited. NC State's policies and regulations covering discrimination,
harassment, and retaliation may be accessed at http://policies.ncsu.edu/policy/pol-04-25-05
or http://www.ncsu.edu/equal_op/.
Any person who feels that he or she has been the subject of prohibited
discrimination, harassment, or retaliation should contact the Office for Equal
Opportunity (OEO) at 919-515-3148.
NOTE: The course schedule is subject to change.
· AI & ML & Deep
Learning; challenges to real-time AI and ML; course info
· Neural Network Architecture
Search, DNN compression overview
· Pruning; Quantization
· Deep Reuse for DNN; DNN
Compilation