Skip to main content

Package for training Gaussian process-like Bayesian Neural Networks with composite structure.

Project description

AutoBNN

This library contains code to specify BNNs that correspond to various useful GP kernels and assemble them into models using operators such as Addition, Multiplication and Changepoint.

It is based on the ideas in the following papers:

  • Lassi Meronen, Martin Trapp, Arno Solin. Periodic Activation Functions Induce Stationarity. NeurIPS 2021.

  • Tim Pearce, Russell Tsuchida, Mohamed Zaki, Alexandra Brintrup, Andy Neely. Expressive Priors in Bayesian Neural Networks: Kernel Combinations and Periodic Functions. UAI 2019.

  • Feras A. Saad, Brian J. Patton, Matthew D. Hoffman, Rif A. Saurous, Vikash K. Mansinghka. Sequential Monte Carlo Learning for Time Series Structure Discovery. ICML 2023.

Setup

AutoBNN has three additional dependencies beyond those used by the core Tensorflow Probability package: flax, scipy and jaxtyping. These can be installed by running setup_autobnn.sh.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

autobnn-0.0.2.dev0.tar.gz (29.4 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page