English

Model Architecture Adaption for Bayesian Neural Networks

Machine Learning 2022-02-10 v1 Artificial Intelligence

Abstract

Bayesian Neural Networks (BNNs) offer a mathematically grounded framework to quantify the uncertainty of model predictions but come with a prohibitive computation cost for both training and inference. In this work, we show a novel network architecture search (NAS) that optimizes BNNs for both accuracy and uncertainty while having a reduced inference latency. Different from canonical NAS that optimizes solely for in-distribution likelihood, the proposed scheme searches for the uncertainty performance using both in- and out-of-distribution data. Our method is able to search for the correct placement of Bayesian layer(s) in a network. In our experiments, the searched models show comparable uncertainty quantification ability and accuracy compared to the state-of-the-art (deep ensemble). In addition, the searched models use only a fraction of the runtime compared to many popular BNN baselines, reducing the inference runtime cost by 2.98×2.98 \times and 2.92×2.92 \times respectively on the CIFAR10 dataset when compared to MCDropout and deep ensemble.

Keywords

Cite

@article{arxiv.2202.04392,
  title  = {Model Architecture Adaption for Bayesian Neural Networks},
  author = {Duo Wang and Yiren Zhao and Ilia Shumailov and Robert Mullins},
  journal= {arXiv preprint arXiv:2202.04392},
  year   = {2022}
}
R2 v1 2026-06-24T09:28:02.866Z