Skip to content

Latest commit

 

History

History
18 lines (11 loc) · 682 Bytes

README.md

File metadata and controls

18 lines (11 loc) · 682 Bytes

Code for the paper Riemannian Laplace Approximation with the Fisher Metric

Accepted to AISTATS 2024

Structure

  • np_jax contains code for experiments other than neural network ones,
  • nn_laplace contains code for neural network experiments.

For further details concerning each refer to the README.md inside each directory.

Minor notes

Some figures are resized to smaller sizes afterwards using sips.

ChatGPT was used as an assistance to write codes. Adopted suggestions, among others, e.g. for writing helper scripts, include

  1. Use metric = 0.5 * (metric + metric.T) partly with suggestion from ChatGPT,
  2. Use torch.square with suggestion from ChatGPT.