The Deep Weight Prior
This work addresses the challenge of data efficiency in training neural networks, particularly for Bayesian models, but it is incremental as it builds on existing prior frameworks.
The paper tackles the problem of incorporating prior knowledge into neural networks by proposing a deep weight prior (DWP) that uses generative models to structure convolutional filters, showing it improves Bayesian neural network performance with limited data and accelerates training of conventional CNNs.
Bayesian inference is known to provide a general framework for incorporating prior knowledge or specific properties into machine learning models via carefully choosing a prior distribution. In this work, we propose a new type of prior distributions for convolutional neural networks, deep weight prior (DWP), that exploit generative models to encourage a specific structure of trained convolutional filters e.g., spatial correlations of weights. We define DWP in the form of an implicit distribution and propose a method for variational inference with such type of implicit priors. In experiments, we show that DWP improves the performance of Bayesian neural networks when training data are limited, and initialization of weights with samples from DWP accelerates training of conventional convolutional neural networks.