Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Improving the Explain-Any-Concept by Introducing Nonlinearity to the Trainable Surrogate Model

Mounes Zaval, Sedat Ozer

TL;DR

By introducing an additional nonlinear layer to the original surrogate model, it is shown that the EAC model can be improved by introducing an additional nonlinear layer to the original surrogate model.

Abstract

In the evolving field of Explainable AI (XAI), interpreting the decisions of deep neural networks (DNNs) in computer vision tasks is an important process. While pixel-based XAI methods focus on identifying significant pixels, existing concept-based XAI methods use pre-defined or human-annotated concepts. The recently proposed Segment Anything Model (SAM) achieved a significant step forward to prepare automatic concept sets via comprehensive instance segmentation. Building upon this, the Explain Any Concept (EAC) model emerged as a flexible method for explaining DNN decisions. EAC model is based on using a surrogate model which has one trainable linear layer to simulate the target model. In this paper, by introducing an additional nonlinear layer to the original surrogate model, we show that we can improve the performance of the EAC model. We compare our proposed approach to the original EAC model and report improvements obtained on both ImageNet and MS COCO datasets.

Improving the Explain-Any-Concept by Introducing Nonlinearity to the Trainable Surrogate Model

TL;DR

By introducing an additional nonlinear layer to the original surrogate model, it is shown that the EAC model can be improved by introducing an additional nonlinear layer to the original surrogate model.

Abstract

In the evolving field of Explainable AI (XAI), interpreting the decisions of deep neural networks (DNNs) in computer vision tasks is an important process. While pixel-based XAI methods focus on identifying significant pixels, existing concept-based XAI methods use pre-defined or human-annotated concepts. The recently proposed Segment Anything Model (SAM) achieved a significant step forward to prepare automatic concept sets via comprehensive instance segmentation. Building upon this, the Explain Any Concept (EAC) model emerged as a flexible method for explaining DNN decisions. EAC model is based on using a surrogate model which has one trainable linear layer to simulate the target model. In this paper, by introducing an additional nonlinear layer to the original surrogate model, we show that we can improve the performance of the EAC model. We compare our proposed approach to the original EAC model and report improvements obtained on both ImageNet and MS COCO datasets.
Paper Structure (5 sections, 4 equations, 2 figures, 2 tables)

This paper contains 5 sections, 4 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Enhanced EAC framework
  4. Experiments
  5. Conclusion

Figures (2)

  • Figure 1: An overview of the explainability process based on the Explain Any Concept (EAC) eac. The top figure shows applying a given classification network on an image and its output. The bottom figure shows the stages of the explainability process including Segment Anything Model (SAM) sam, the surrogate model (our version) and the Shapley values.
  • Figure 2: (a) shows the PIE for the original explain-any-concept (EAC) as described in eac. (b) shows our proposed PIE which includes a non-linear surrogate model.