Comprehensive Information Bottleneck for Unveiling Universal Attribution to Interpret Vision Transformers

TL;DR

This paper introduces CoIBA, a comprehensive information bottleneck method that explains vision transformer decisions by capturing relevant information across multiple layers, improving attribution faithfulness.

Contribution

It proposes a novel multi-layer information bottleneck approach with shared damping ratio to better explain decision processes in vision transformers.

Findings

Enhanced attribution faithfulness demonstrated in experiments

Effective discovery of decision-relevant information across layers

Shared damping ratio improves information sharing across layers

Abstract

The feature attribution method reveals the contribution of input variables to the decision-making process to provide an attribution map for explanation. Existing methods grounded on the information bottleneck principle compute information in a specific layer to obtain attributions, compressing the features by injecting noise via a parametric damping ratio. However, the attribution obtained in a specific layer neglects evidence of the decision-making process distributed across layers. In this paper, we introduce a comprehensive information bottleneck (CoIBA), which discovers the relevant information in each targeted layer to explain the decision-making process. Our core idea is applying information bottleneck in multiple targeted layers to estimate the comprehensive information by sharing a parametric damping ratio across the layers. Leveraging this shared ratio complements the over-compressed information to discover the omitted clues of the decision by sharing the relevant information across the targeted layers. We suggest the variational approach to fairly reflect the relevant information of each layer by upper bounding layer-wise information. Therefore, CoIBA guarantees that the discarded activation is unnecessary in every targeted layer to make a decision. The extensive experimental results demonstrate the enhancement in faithfulness of the feature attributions provided by CoIBA.