Research map / Learning theory and algorithms / Prediction and decision losses
Axiomatic Attribution for Deep Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Integrate gradients along the straight-line path from the baseline to the input instead of using gradients at the input | ImageNet object recognition model · ImageNet object recognition images | visualization quality | Helped |
| Use integrated gradients rather than gradients at the input | Deep-network attribution · Deep-network attribution | Sensitivity(a) | Helped |
| Use integrated gradients rather than DeepLift | Deep-network attribution · Functionally equivalent networks | Implementation Invariance | Helped |
| Use integrated gradients rather than LRP | Deep-network attribution · Functionally equivalent networks | Implementation Invariance | Helped |
| Use integrated gradients rather than Deconvolution | Deep-network attribution · Network with output sensitive to x2 | Sensitivity(a) | Helped |
| Use integrated gradients rather than Guided back-propagation | Deep-network attribution · Network with output sensitive to x2 | Sensitivity(a) | Helped |
| Use a straight-line path rather than another path | Path-integrated attribution methods · Symmetric input variables | symmetry preservation | Helped |
| Use integrated gradients rather than attention weights alone | LSTM-based neural machine translation · Input-output token alignment | attribution completeness | Helped |
Closest papers by meaning
- Fast Axiomatic Attribution for Neural Networks (NeurIPS 2021 · similarity 0.92)
- Towards better understanding of gradient-based attribution methods for Deep Neural Networks (ICLR 2018 · similarity 0.91)
- MFABA: A More Faithful and Accelerated Boundary-Based Attribution Method for Deep Neural Networks (AAAI 2024 · similarity 0.89)
- Iterative Search Attribution for Deep Neural Networks (ICML 2024 · similarity 0.87)
- How to Probe: Simple Yet Effective Techniques for Improving Post-hoc Explanations (ICLR 2025 · similarity 0.86)
- Benchmarking the Attribution Quality of Vision Models (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.