GSoC 2022 Summary
This summer I participated in Google Summer of Code 2022 as a contributor, and I worked with Kevin P. Murphy on the PyProbML repo, which is a complement to the textbook “Probabilistic Machine Learning: Advanced Topics”.
My work can be roughly separated into two parts. In the first half of the summer, I worked on adding demos and short sections to the textbook. In the second half, I joined the ssm-jax group and mainly focused on contributing to the ssm-jax package.
Highlights
1. Work on the text book
-
Add demos of Dirichlet process (DP) mixture model for clustering
-
Add a subsection on Variational inference of DP mixture model.
2. Work on the ssm-jax package
-
Add subsections on learning and inference of state space models to the textbook.
-
Add the class Linear Gaussian state space model (LGSSM) to the ssm-jax package. This model is a general framework which takes structural time series models as special cases and can be further used for causal inference.
-
Add LGSSM models with conjugate priors to the ssm-jax package.
Summary
I am honored to work with Kevin this summer to contribute to the textbook the orginal version of which has influenced me a lot. The advisors in the team of ssm-jax are very supportive and always willing to help. It is also a wonderful experience to collaborate with all members in this project. I am inspired by the their enthusiam for the the product we are making together.