bayne 0.1.1

Bayesian Neural Networks in Pytorch

Bayne - Bayesian Neural Networks

Pytorch/Pyro-based Bayesian Neural Networks library

Features:

• Markov Chain Monte Carlo (MCMC)
  • Full connectected neural network with ReLU, Sigmoid, or Tanh activation
  • Configurable prior
  • Hamiltonian Monte Carlo or NUTS sampler
• Bounds propagation
  • CROWN - linear bounds
  • IBP - interval bounds
  • CROWN-IBP - linear bounds

Authors:

• Frederik Baymler Mathiesen, PhD student @ TU Delft, Netherlands