Bayesian Learning of Neural Network Architectures

2019/01/14 by Georgi Dikov, Dikov, Georgi, Patrick van der Smagt +3
Computer Science · Mathematics · #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.1901.04436

The 22nd International Conference on Artificial Intelligence and Statistics (AISTATS 2019)

arxiv created 2019/01/27 · arxiv updated 2019/01/29

Abstract

In this paper we propose a Bayesian method for estimating architectural parameters of neural networks, namely layer size and network depth. We do this by learning concrete distributions over these parameters. Our results show that regular networks with a learnt structure can generalise better on small datasets, while fully stochastic networks can be more robust to parameter initialisation. The proposed method relies on standard neural variational learning and, unlike randomised architecture search, does not require a retraining of the model, thus keeping the computational overhead at minimum.

Related

• Measuring Uncertainty through Bayesian Learning of Deep Neural Network Structure
• Optimisation & Generalisation in Networks of Neurons
• Bayesian Learning for Neural Networks: an algorithmic survey
• Learning Neural Network Architectures using Backpropagation
• Building Bayesian Neural Networks with Blocks: On Structure, Interpretability and Uncertainty