Practical Bayesian Learning of Neural Networks via Adaptive Optimisation Methods
This work addresses the need for reliable uncertainty quantification in neural networks, which is crucial for applications like decision-making under uncertainty, though it appears incremental as it builds on existing optimization methods.
The authors tackled the problem of estimating posterior distributions over neural network weights by introducing a probabilistic interpretation of adaptive optimization algorithms like AdaGrad and Adam, resulting in a method called Badam that effectively learns uncertainties for tasks such as weight pruning and Thompson sampling in multi-armed bandits.
We introduce a novel framework for the estimation of the posterior distribution over the weights of a neural network, based on a new probabilistic interpretation of adaptive optimisation algorithms such as AdaGrad and Adam. We demonstrate the effectiveness of our Bayesian Adam method, Badam, by experimentally showing that the learnt uncertainties correctly relate to the weights' predictive capabilities by weight pruning. We also demonstrate the quality of the derived uncertainty measures by comparing the performance of Badam to standard methods in a Thompson sampling setting for multi-armed bandits, where good uncertainty measures are required for an agent to balance exploration and exploitation.