Latent variable models (LVMs) are powerful and flexible tools for learning hidden structure underlying data objects in an unsupervised fashion. They provide a compact, meaningful representation of the inputs. Bayesian matrix factorization is a general class of LVMs which factorizes a data matrix into a product of two low-rank matrices. LVMs (and BMF) can be used for many purposes in machine learning, such as clustering, pattern recognition, dimensionality reduction, feature extraction, and predicting missing values.
Latent variable modeling is a very broad research topic. This seminar course will provide a gateway to this interesting topic through an introductory lecture for several widely used LVMs (e.g. factor analysis and multi-view learning) and basic principles of the methods as well as discussion of more recent advances in the field.
The course will include (1) an introductory lecture about the main topics, (2) student presentations of selected readings of research papers for an overview of the topics, and optionally (3) a project work for hands-on experience and a deeper understanding of the topics.
The project work will be based on pen & papers and R/Python & STAN programming language.
The presentation should explain the background / motivation, major concept of the selected papers, and possibly the connection with the other / previous presentation, etc. Everyone is expected to be familiar with the discussed papers prior to the seminar. Following each presentation there will be a discussion with all course participants about the contributions of the papers and the questions remaining open. Active participation is strongly encouraged.
The course is mainly aimed at doctoral students and advanced master's students. Note that due to the format of the course, the number of students is limited to maximum 15.
Basic Mathematics, familiarity with Machine Learning basic principles is a plus.
Period V, 13.04.2018 - 18.05.2018.
Lectures in R030/A133 T5, Fridays, 10:15 - 12:00.
3 ECTS for presenting papers and active participation
5 ECTS, requirement for 3 ECTS + doing one of the proposed projects
|1. Do one presentation.||Grade points|| 3 ECTS
|| 5 ECTS
|2. Lead discussion for one presentation.||1||Required||Required|
|3. Submit 3 reading diaries on papers presented by others.||1||Required||Required|
|Get one grade point per learning diary.||0 - 3
|4. Finish the proposed project and write a project report.||2||Required|
- For 3 ECTS: grade=min(5, 2+number of learning diaries returned); 0 if no paper presentation done and not leading discussion for any presentation; all learning diaries must be returned to mycourses before the specific deadlines to count.
- For 5 ECTS: grade=(2/5) * grade of project work + (3/5) * min(5, 2+number of learning diaries returned); 0 if no paper presentation done (and 3 ECTS if no project work returned); all course work must be returned to mycourses by the end of May, 31.5.2018, to count.
- Task 1 and 2 required for 3 ECTS
- Task 1, task 2 and task 4 required for 5 ECTS
Xiangju Qin, Paul Blomstedt; Probabilistic Machine Learning research group
Advisor: Prof. Samuel Kaski