/probprog20

Primary LanguageClojure

CS423 Probabilistic Programming, Spring 2020, KAIST

This is a webpage of the course "CS423 Probabilistic Programming", which is offered at the KAIST CS department in the spring of 2020. The webpage will contain links to lecture slides and other course-related materials.

Probabilistic programming refers to the idea of developing a programming language for writing and reasoning about probabilistic models from machine learning and statistics. Such a language comes with the implementation of several generic inference algorithms that answer various queries about the models written in the language, such as posterior inference and marginalisation. By providing these algorithms, a probabilistic programming language enables data scientists to focus on designing good models based on their domain knowledge, instead of building effective inference engines for their models, a task that typically requires expertise in machine learning, statistics and systems. Even experts in machine learning and statistics may get benefited from such a probabilistic programming system because using the system they can easily explore highly advanced models.

This course has two goals. The first is to help students to be a good user of an expressive probabilistic programming language. Throughout the course, we will use a particular language, called Anglican, but we will emphasise general principles that apply to a wide range of existing probabilistic programming systems. The second goal is to expose the students to recent exciting results in probabilistic programming, which come from machine learning, statistics, programming languages, and probability theory.

1. Important Announcements

[June 8] Group presentations on 24 June

Because of a special circumstance of some students, we will have to change the order of group project presentations on the 24th of June. The new order is:

  • [Group 14] Seokin Seo, Jinu Choi, Juhyung Park
  • [Group 11] Dongjoon Yun
  • [Group 12] Euojin Kim
  • [Group 13] Daeseok Lee

We hope that those who get affected understand it.

[June 7] Information about final exam

The final exam will be take-home. The exam questions will become available from KLMS at 10am on 29 June (Monday). You will have to submit your answers in KLMS by 10pm on the same date. That is, you will be given 12 hours. It is ok to look at the materials from KLMS and the course webpage: lecture notes/slides, recorded lectures in KLMS, Q&A in KLMS, and homework sheets. That is, you can use anything from the KLMS page and the course webpage. But you shouldn't use anything beyond. For instance, you should not use the online materials linked to in the course webpage. If you are found to violate the honour code (for instance, by exchanging information with your friends or googling information), you will get F automatically; we may also report the incident to the department and the university.

Because of some fairness issue, we might be unable to answer questions during the exam. We will try to make exam questions as clear as possible so that no questions really arise. However, if you find that something is ambiguous during the exam and we cannot answer your questions, we advise you to make assumptions and write down those assumptions explicitly in your answers.

[June 7] Group project presentation on 17/22/24 June

The last three lectures of the course will consist of project presentations. Each group will have 15 minutes to explain what they did and experienced during their project. The groups 1-5 will present on 17 June, the groups 6-10 on 22 June, and the remaining groups 11-14 on 24 June. In addition to this group presentation, each group should write a short report (max 3 pages) summarising what they did and learnt during the group project, and submit the report in KLMS. The deadline for the project report is 6:00pm on 25 June (Friday). This is a hard deadline; late submissions are not allowed. One report is enough for each group.

  • 17 June (Wednesday)
  1. Herin Park, Hyojin Shim
  2. Eunseob Kim, Sumin Shin
  3. Taeyoung Kim, Janggun Lee, Geon Park
  4. Gabin An, Sungmin Kang
  • 22 June (Monday)
  1. Jaemin Kim, Yunseok Lee
  2. Tung Duong Mai, Assem Zhunis
  3. Euicheon Lee, Hyunjae Lee
  4. Hongchan An
  • 24 June (Wednesday)
  1. Seokin Seo, Jinu Choi, Juhyung Park
  2. Dongjoon Yun
  3. Euojin Kim
  4. Daeseok Lee
  5. Kwangjo Gong, Donghyeon Kim

[May 31] Homework4 is out.

The homework assignment 4 is out. Submit your solutions in KLMS by 6:00pm on 17 June 2020 (Wednesday).

[May 9] Homework3 is out.

The homework assignment 3 is out. Submit your solutions in KLMS by 6:00pm on 29 May 2020 (Friday).

[April 19] Group project pitch on 27/29 April

On the 27 and 29 of April, we will have project-pitch sessions in our ZOOM online classes. This means that each group will have 10 minutes to explain what they plan to do for their group project and to get feedback from me and other participants. Here is the schedule for project-pitch presentations in those two days.

  • 27 April (Monday)
  1. Kwangjo Gong, Donghyeon Kim
  2. Herin Park, Hyojin Shim
  3. Eunseob Kim, Sumin Shin
  4. Taeyoung Kim, Janggun Lee, Geon Park
  5. Gabin An, Sungmin Kang
  6. Jaemin Kim, Yunseok Lee
  7. Tung Duong Mai, Assem Zhunis
  • 29 April (Wednesday)
  1. Taehyouk Cho
  2. Euicheon Lee, Hyunjae Lee
  3. Hongchan An
  4. Dongjoon Yun
  5. Euojin Kim
  6. Daeseok Lee
  7. Seokin Seo, Jinu Choi

So, each group should prepare a 8-min presentation on what they plan to do on the project. In the presentation, they can explain what problem or challenge related to probabilistic programming they want to solve, why they want to do this, how they plan to attack the problem or challenge using probabilistic programming systems, and what others have done in that domain. But this is just my suggestion, and each group may pick the format of presentation. Think of this as a good opportunity to get feedback from your friends in the class.

Doing this presentation is a part of your project. So, don't miss it. Although we strongly encourage you to present it in English, it is acceptable to give a talk in Korean. But even in that case, make sure that your slides are written in English, so that non-native Korean speakers at least have some idea about what goes on.

It is ok to change your plan later. It is also ok that your project may fail to achieve the original goal even if you work hard on it. When we mark your project, we will consider all of these.

[April 12] Homework2 is out.

The homework assignment 2 is out. Submit your solutions in KLMS by 6:00pm on 22 April 2020 (Wednesday).

[April 8] No lecture on 15 April.

The 15th of April is an election day in Korea. We won't have an online class on that day.

[March 28] Homework1 is out.

The homework assignment 1 is out. Submit your solutions in KLMS by 6:00pm on 8 April 2020 (Wednesday).

We remind the students that we adopt a very strict policy for handling dishonest behaviours. If a student is found to copy answers from peers or other sources in her or his submission for this homework assignment, he or she will get F.

Best wishes, Hongseok

[March 14] Penality scheme for late homework submissions.

Here is our scheme for handling late homework submissions. It assumes that the total marks are 100.

  1. <= One day late: -10
  2. <= Two days late: -20
  3. <= Three days late: -30
  4. <= Four days late: -40
  5. More than four days late: -100.

[March 14] Project group.

The group project is an important part of this course. Find your project partners by 30 March 2020 (Monday), and inform all of Hongseok, Daeseok, and Kwonsoo by email by the midnight of 30 March 2020. Each group should consist of 1-3 students. (We allow a group of one student because of the special circumstance of moving a large amount of course activities online this year.)

[March 8] Honour code.

The course will adopt very strict honour code. If a student is found to copy homework solutions from friends and other sources, he or she will get F. Similarly, if a student is found to do cheaing during exam, she or he will get F as well.

[March 8] Online lectures for the first two weeks at least.

Because of Covid-19, the lectures for the first two weeks at least (16 March 2020 - 27 March 2020) will be held online at the usual schedule. We will use ZOOM, and you should have received information about how to access it from the university and the department. The access keys to online sessions will be provided in KLMS.

Another measure that we take is not to have any face-to-face meetings between teaching staffs and students, again because of Covid-19. This means that offline office hours won't be held for a while. To get help, please use the discussion forums in KLMS.

2. Logistics

Evaluation

  • Final exam (40%). Project (40%). Homework (20%).

Teaching Staffs

  • Lecturer: Prof Hongseok Yang (email: hongseok00@gmail.com, office hours: 6:00pm - 7:00pm on Monday in ZOOM. You can find more information about it in KLMS.)
  • TA1: Gwonsoo Che (email: gwonsoo.che@gmail.com, office hours: 6:00pm - 7:00pm every other Thursday through ZOOM. See KLMS announcements for more information.)
  • TA2: Daeseok Lee (email: daeseoklee@kaist.ac.kr, office hours: 7:30pm - 8:30pm every other Thursday through ZOOM. See KLMS announcements for more information.)

Place and Time

  • Place: room 111 in the N1 building
  • Time: 1:00pm - 2:15pm on Monday and Wednesday from 16 March 2020 until 29 June 2020. For the first weeks and possibly more, we will have virtual online classes using ZOOM.
  • Final exam: 1:00pm - 3:00pm on 10 June 2019 (Monday) at the room 111 in the N1 building.

Online Discussion

  • We will use KLMS.

3. Homework

Submit your solutions in KLMS. We will create submission folders for all the homework assignments in KLMS.

4. Tentative Plan

  • 03/16 (Mon) - Introduction. Slides. Homework0. Gorilla worksheet
  • 03/18 (Wed) - Basics of Clojure and Tiny Bit of Anglican. Slides. Gorilla worksheet. Programs.
  • 03/23 (Mon) - Basics of Clojure and Tiny Bit of Anglican.
  • 03/25 (Wed) - Posterior Inference, Basics of Anglican, and Importance Sampling. Slides.
  • 03/30 (Mon) - Posterior Inference, Basics of Anglican, and Importance Sampling.
  • 04/01 (Wed) - Generative Modelling with Anglican. Slides. Gorilla worksheet for 2D physics. Gorilla worksheet for program induction. (Note: In order to run the 2D physics example, you will have to copy bounce.clj to the anglican-user/src directory and replace anglican-user/project.clj by this file).
  • 04/06 (Mon) - Markov Chain Monte Carlo. Slides.
  • 04/08 (Wed) - Markov Chain Monte Carlo.
  • 04/13 (Mon) - Annealed Importance Sampling. Neal's original paper. Copies of virtual whiteboards (Note1, Note2, Note3).
  • 04/15 (Wed) - NO LECTURE. Election Day.
  • 04/20 (Mon) - Implementing Inference Algorithms for Probabilistic Programs. Slides. Note. Handwritten Notes: (1, 2, 3, 4, 5, 6, 7, 8, 9, 10). Copies of virtual whiteboards (Note1, Note2, Note3).
  • 04/22 (Wed) - Implementing Inference Algorithms for Probabilistic Programs.
  • 04/27 (Mon) - Group Project Pitch.
  • 04/29 (Wed) - Group Project Pitch.
  • 05/04 (Mon), 05/06 (Wed) - NO LECTURES. Midterm Exam.
  • 05/11 (Mon) - Implementing Inference Algorithms for Probabilistic Programs.
  • 05/13 (Wed) - Implementing Inference Algorithms for Probabilistic Programs. Copies of virtual whiteboards (Note1, Note2, Note3, Note4, Note5).
  • 05/18 (Mon) - Stochastic Variational Inference. Slides. Handwritten Notes: (1, 2, 3). Correction: 1.
  • 05/20 (Wed) - NO LECTURE. Graduate Admission.
  • 05/25 (Mon) - Stochastic Variational Inference.
  • 05/27 (Wed) - Amortised Inference. Slides.
  • 06/01 (Mon) - Reparameterisation Gradient, Model Learning via ELBO, IWAE and Doubly-Reparameterised Gradient Estimator. Handwritten Notes: 1, 2, 3, 4. Incomprehensible Handwritten Notes on Doubly-Reparameterised Gradient Estimator: 1, 2, 3.
  • 06/03 (Wed) - Reparameterisation Gradient, Model Learning via ELBO, IWAE and Doubly-Reparameterised Gradient Estimator.
  • 06/08 (Mon) - Denotational Semantics of Probabilistic Programs. Slides. Note.
  • 06/10 (Wed) - Denotational Semantics of Probabilistic Programs.
  • 06/15 (Mon) - Denotational Semantics of Probabilistic Programs.
  • 06/17 (Wed) - Student Presentation.
  • 06/22 (Mon) - Student Presentation.
  • 06/24 (Wed) - Student Presentation.
  • 06/29 (Mon), 07/01 (Wed) - NO LECTURES. Final Exam.

5. Studying Materials

Studying the lecture slides and notes and the homework exercises of the course is likely to be the most time-efficient way to catch up with this course. Also, at each lecture, we will give students pointers to the related papers. If a student does not understand a certain concept, we encourage him or her to look it up in the Internet. We typically do this when we encounter a similar problem. In our case, Wikipedia, lecture notes or survey articles have helped us the most.

The next best option is to read the following draft book on probabilistic programming:

  1. "An Introduction to Probabilistic Programming" by Jan-Willem van de Meent, Brooks Paige, Hongseok Yang and Frank Wood. Reading this book will give a broader view on probabilistic programming and much deeper understanding into its inference algorithms and their implementations.

If a student feels that she or he lacks background knowledge on machine learning, we recommend him or her to have a look at the following online materials.

  1. The online book "Probabilistic Programming and Bayesian Methods for Hackers" describes Bayesian Machine Learning using a probabilistic programming system called PyMC. Hongseok found this book easy to follow and good at explaining basics and intuitions.

  2. A more standard reference on machine learning is Bishop's book "Pattern Recognition and Machine Learning".

Two good ways to understand probabilistic programming are to try a wide range of examples and to understand common implementation techniques for probabilistic programming languages. The following documents provide such examples or explain those techniques.

  1. Anglican website. In particular, students will learn a lot by trying examples in the site.

  2. Forestdb.org is a great source of interesting probabilistic programs.

  3. Edward tutorial website and Pyro example website. Edward and Pyro are so called deep probabilistic programming languages that attempt to combine deep learning and probabilistic programming. These web pages contain interesting examples that one can try using these languages.

  4. Goodman and Stuhlmuller's book "The Design and Implementation of Probabilistic Programming Languages". This web-based book describes the implementation of WebPPL, a probabilistic programming language on top of JavaScript. Many techniques in the book are general and apply to other probabilistic programming languages.

Frank Wood taught a graduate-level course on probabilistic programming at UBC. It emphasises on actually building efficient runtimes for probabilistic programming languages. Here is a link to his course.

  1. Wood's graduate-level course at UBC.

6. Group Project

  1. Kwangjo Gong, Donghyeon Kim
  2. Herin Park, Hyojin Shim
  3. Eunseob Kim, Sumin Shin
  4. Taeyoung Kim, Janggun Lee, Geon Park
  5. Gabin An, Sungmin Kang
  6. Jaemin Kim, Yunseok Lee
  7. Tung Duong Mai, Assem Zhunis
  8. Taehyouk Cho
  9. Euicheon Lee, Hyunjae Lee
  10. Hongchan An
  11. Dongjoon Yun
  12. Euojin Kim
  13. Daeseok Lee
  14. Seokin Seo, Jinu Choi, Juhyung Park