Website "deep2Read" for UVA Qdata Group's Deep Learning Journal Club
– By [Dr. Yanjun Qi] @https://www.cs.virginia.edu/yanjun/
About this website:
- As a group, we need to improve our knowledge of the fast-growing field of deep learning
- To educate students in our graduate programs, to help new members in my team with basic tutorials, and to help current members understand advanced topics better, this website includes a (growing) list of tutorials and papers we survey for such a purpose.
- We hope this website helps people who share similar research interests or those interested in learning advanced topics about deep learning.
- Please feel free to email me (yanjun@virginia.edu), if you have comments, questions or recommendations.
About the Sidebar and how we group the readings
-
Until Dec. 2020, we have shared reviewes slides for about 500 deep learning papers via this web site.
-
Until Dec. 2020, we have finished 92 reading sessions. Each reading session reviewed multiple deep learning papers. Each post in this website summarizes one of our reading session.
-
Due to the large number of slides, we try to organize them according to three different variables. Each variable maps to one item in this website’s top header bar.
-
ByCategory: we group our readings to 10 general groups: 0Basics , 1Theoretical , 2Architecture , 2Graphs , 3Reliable , 4Optimization , 5Generative , 6Reinforcement , 7MetaDomain , 8Scalable , 9DiscreteApp . Please check out these categories and their relevant readings via Readings ByCategory . As shortcuts, we also add each category’s page URL as a single item (sorted by Category Names) in our side bar.
-
ByTags: we use 150 different tags to organize deep learning papers we reviewed. Please check out these tags via Readings ByTag .
-
ByReadDates: our review slides were done in multiple different semesters (2017Fall, 2018Club, 2019Spring, 2019Fall). To help students (from a specific semester) navigate, we also group our reading sessions according to what semester a survey talk happened at (via our seminar courses or via our reading club). Please check out our readings sorted by semesters via Readings ByReadDate
Claim
- The covered tutorials and papers are by no means an exhaustive list, but are topics which we have learned or plan to learn.
History
- This website was started from two seminar courses I taught at UVA in Fall 2017 and Spring 2019. Later I expand the content with my team reaing group
- The seminar courses offer opportunities for students to have in-depth understanding and hands-on experience of deep learning. Students are expected to generate top-tier publications when finishing the course.
- The materials aim to offer opportunities for students to have in-depth understanding and hands-on experience of advances in deep learning.
Course Basics and General Description
(The following content is for when we have the seminar course)
- This is an advanced graduate-level deep learning course.
- The course takes the form of half-seminar and half-project. The form of seminar focuses on paper readings.
- This course offers opportunities for students to get into research topics about the state-of-the-art advanced deep learning.
- No text book
- Sit-in: No. This course is for registered students only.
Instructor
- Prof. Yanjun Qi
- EMail: yanjun@virginia.edu
- Rice Hall 503 , 243-3089
- Office hours: Wed 9:30am-12:30pm.
Prerequisite:
- Instructor’s Permission for enrollment is required for this course.
- Required courses as prerequisite: Graduate-level machine learning; Introduction of Deep Learning and Graduate-level Optimization are preferred.
- Familiar reading of Basic Deep Learning are preferred.
Course Grading Policy
The grade will be calculated as follows:
- 60% for the in-class paper presentations/discussions/ note taking
- 40% for the project
Assignments
- Sharelatex/overleaf to submit lectures about the assigned papers
-
Each class, we will assign 4 to 6 reading materials (video lectures or papers or research lecture slides )
-
Each student is expected to have three sets of assigntments:
- (a) Weekly project summary should be updated per week right before project meetings with Prof. Qi;
- (b) Assigned presentation slides: please use the BEAMER template shared through the course overleaf project. Please make sure the presentation slides are ready before every Friday 8am; (One Example Slide Presenation)
-
(c) Assigned scribe notes: please use the latex template shared through the course overleaf project. Please make sure the scribe notes are ready one week after. (One Example Scribe note)
- For both the paper presentations and the scribe notes, please use the following structure as reference:
- Full reference of the paper
- Motivations / Why needed ? / Why important ?
- Previous solutions
- Key insights
- Key equations
- Key conclusions
- Goals achieved: / Under what restrictions or assumptions;
Logistics Information
- Announcements are being emailed to the course mailing list.
- A welcome note will be sent to the mailing list early in the semester.
- Errata and answers to questions are being discussed and answered on the course emailist.