Bayesian Neural Networks Avoid Encoding Complex and Perturbation-Sensitive Concepts
This addresses the problem of improving robustness and generalization in neural networks for AI practitioners, though it is incremental as it builds on prior work on concept encoding.
The paper proves that Bayesian Neural Networks (BNNs) are less likely to encode complex and perturbation-sensitive concepts compared to standard deep neural networks, with experiments verifying this theoretical finding.
In this paper, we focus on mean-field variational Bayesian Neural Networks (BNNs) and explore the representation capacity of such BNNs by investigating which types of concepts are less likely to be encoded by the BNN. It has been observed and studied that a relatively small set of interactive concepts usually emerge in the knowledge representation of a sufficiently-trained neural network, and such concepts can faithfully explain the network output. Based on this, our study proves that compared to standard deep neural networks (DNNs), it is less likely for BNNs to encode complex concepts. Experiments verify our theoretical proofs. Note that the tendency to encode less complex concepts does not necessarily imply weak representation power, considering that complex concepts exhibit low generalization power and high adversarial vulnerability. The code is available at https://github.com/sjtu-xai-lab/BNN-concepts.