Explainable embeddings with Distance Explainer

TL;DR

Distance Explainer provides a novel, post-hoc method for interpreting the meaning of distances in embedded vector spaces, improving transparency in models like CLIP and cross-modal embeddings.

Contribution

It introduces a new explanation technique for embedded spaces that adapts saliency methods to explain pairwise distances, filling a gap in XAI for vector representations.

Findings

Effectively identifies features influencing similarity/dissimilarity in embeddings

Maintains high robustness and consistency in explanations

Parameter tuning impacts explanation quality

Abstract

While eXplainable AI (XAI) has advanced significantly, few methods address interpretability in embedded vector spaces where dimensions represent complex abstractions. We introduce Distance Explainer, a novel method for generating local, post-hoc explanations of embedded spaces in machine learning models. Our approach adapts saliency-based techniques from RISE to explain the distance between two embedded data points by assigning attribution values through selective masking and distance-ranked mask filtering. We evaluate Distance Explainer on cross-modal embeddings (image-image and image-caption pairs) using established XAI metrics including Faithfulness, Sensitivity/Robustness, and Randomization. Experiments with ImageNet and CLIP models demonstrate that our method effectively identifies features contributing to similarity or dissimilarity between embedded data points while maintaining high robustness and consistency. We also explore how parameter tuning, particularly mask quantity and selection strategy, affects explanation quality. This work addresses a critical gap in XAI research and enhances transparency and trustworthiness in deep learning applications utilizing embedded spaces.