Skip to content

Week 6 and 7 into GSoC

All these days I have experimenting with Full Rank ADVI testing on rugby, radon notebooks and park_bias_model. Now, I am planning to draw general conclusions in what scenarios Variational Inference leads to Cholesky Decomposition errors when the model contains Gaussian Processes. Its mainly due to dtypes and high learning rates.

Gist - Source

Also, I have written a short proposal to explore ways of adding transforms to PyMC4.

Thank you for reading!

With ❤, Sayam

Last update: July 26, 2020