Model and Feature Diversity for Bayesian Neural Networks in Mutual Learning
This work addresses performance issues in BNNs for uncertainty quantification, offering an incremental improvement over existing mutual learning methods.
The paper tackled the underperformance of Bayesian Neural Networks (BNNs) compared to deterministic networks by enhancing mutual learning through increased diversity in parameter and feature distributions, resulting in significant improvements in classification accuracy, negative log-likelihood, and expected calibration error.
Bayesian Neural Networks (BNNs) offer probability distributions for model parameters, enabling uncertainty quantification in predictions. However, they often underperform compared to deterministic neural networks. Utilizing mutual learning can effectively enhance the performance of peer BNNs. In this paper, we propose a novel approach to improve BNNs performance through deep mutual learning. The proposed approaches aim to increase diversity in both network parameter distributions and feature distributions, promoting peer networks to acquire distinct features that capture different characteristics of the input, which enhances the effectiveness of mutual learning. Experimental results demonstrate significant improvements in the classification accuracy, negative log-likelihood, and expected calibration error when compared to traditional mutual learning for BNNs.