mlsanityData to 14 Oct 2026

Unifying Perplexing Behaviors in Modified BP Attributions through Alignment Perspective

arXiv 2503.11160 ↗First posted 14 Mar 2025cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Attributions aim to identify input pixels that are relevant to the decision-making process. A popular approach involves using modified backpropagation (BP) rules to reverse decisions, which improves interpretability compared to the original gradients. However, these methods lack a solid theoretical foundation and exhibit perplexing behaviors, such as reduced sensitivity to parameter randomization, raising concerns about their reliability and highlighting the need for theoretical justification. In this work, we present a unified theoretical framework for methods like GBP, RectGrad, LRP, and DTD, demonstrating that they achieve input alignment by combining the weights of activated neurons. This alignment improves the visualization quality and reduces sensitivity to weight randomization. Our contributions include: (1) Providing a unified explanation for multiple behaviors, rather than focusing on just one. (2) Accurately predicting novel behaviors. (3) Offering insights into decision-making processes, including layer-wise information changes and the relationship between attributions and model decisions.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 34 of its 41 references that are in the map. First posted on arXiv 14 Mar 2025.

Builds on

34 of 41 references.

References · 34

  1. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
  2. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
  3. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  4. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,371
  5. 2013Visualizing and Understanding Convolutional Networkscited by 2,290
  6. 2020Shortcut Learning in Deep Neural Networkscited by 1,969
  7. 2018Sanity Checks for Saliency Mapscited by 1,156
  8. 2017BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chaincited by 1,106
  9. 2014Striving for Simplicity: The All Convolutional Netcited by 990
  10. 2018Robustness May Be at Odds with Accuracycited by 880
  11. 2015Evaluating the visualization of what a Deep Neural Network has learnedcited by 419
  12. 2019A Survey on Explainable Artificial Intelligence (XAI): Towards Medical XAIcited by 404
22 more
  1. 2019Unmasking Clever Hans Predictors and Assessing What Machines Really Learncited by 365
  2. 2016Top-down Neural Attention by Excitation Backpropcited by 321
  3. 2020Input-Aware Dynamic Backdoor Attackcited by 313
  4. 2020A Survey on Neural Network Interpretabilitycited by 291
  5. 2015Explaining NonLinear Classification Decisions with Deep Taylor Decompositioncited by 250
  6. 2020Evaluating and Aggregating Feature-based Model Explanationscited by 188
  7. 2020Dynamic Backdoor Attacks Against Machine Learning Modelscited by 151
  8. 2019Interpretations are useful: penalizing explanations to align neural networks with prior knowledgecited by 145
  9. 2019When Explanations Lie: Why Many Modified BP Attributions Failcited by 92
  10. 2019On the Connection Between Adversarial Robustness and Saliency Map Interpretabilitycited by 80
  11. 2020Attribution in Scale and Spacecited by 61
  12. 2018A Theoretical Explanation for Perplexing Behaviors of Backpropagation-based Visualizationscited by 60
  13. 2020There and Back Again: Revisiting Backpropagation Saliency Methodscited by 60
  14. 2018Local Explanation Methods for Deep Neural Networks Lack Sensitivity to Parameter Valuescited by 40
  15. 2019Why are Saliency Maps Noisy? Cause of and Solution to Noisy Saliency Mapscited by 34
  16. 2020What Do You See? Evaluation of Explainable Artificial Intelligence (XAI) Interpretability through Neural Backdoorscited by 32
  17. 2019Bridging Adversarial Robustness and Gradient Interpretabilitycited by 22
  18. 2022Shortcomings of Top-Down Randomization-Based Sanity Checks for Evaluations of Deep Neural Network Explanationscited by 22
  19. 2021Revisiting Sanity Checks for Saliency Mapscited by 20
  20. 2020A Unified Taylor Framework for Revisiting Attribution Methodscited by 13
  21. 2021Investigating sanity checks for saliency maps with image and text classificationcited by 11
  22. 2024Sanity Checks Revisited: An Exploration to Repair the Model Parameter Randomisation Testcited by 9

Built on by

Not cited yet.