TensorBNN: Bayesian Inference for Neural Networks using Tensorflow

TL;DR

TensorBNN introduces a TensorFlow-based package that performs Bayesian inference on neural networks using Hamiltonian Monte Carlo, efficiently utilizing GPU acceleration for training and prediction.

Contribution

It provides a novel, GPU-accelerated implementation of Bayesian neural networks within TensorFlow, enabling scalable posterior sampling.

Findings

Efficient posterior sampling of neural network parameters.

Seamless integration with TensorFlow's GPU capabilities.

Enhanced uncertainty quantification in neural network predictions.

Abstract

TensorBNN is a new package based on TensorFlow that implements Bayesian inference for modern neural network models. The posterior density of neural network model parameters is represented as a point cloud sampled using Hamiltonian Monte Carlo. The TensorBNN package leverages TensorFlow's architecture and training features as well as its ability to use modern graphics processing units (GPU) in both the training and prediction stages.