Leveraging Sparse Linear Layers for Debuggable Deep Networks
arXiv:2105.04857
Abstract
We show how fitting sparse linear models over learned deep feature representations can lead to more debuggable neural networks. These networks remain highly accurate while also being more amenable to human interpretation, as we demonstrate quantiatively via numerical and human experiments. We further illustrate how the resulting sparse explanations can help to identify spurious correlations, explain misclassifications, and diagnose model biases in vision and language tasks. The code for our toolkit can be found at https://github.com/madrylab/debuggabledeepnetworks.
References in corpus (28)
- Discussion of "Least angle regression" by Efron et al
- Rejoinder to "Least angle regression" by Efron et al
- Least Angle Regression
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Axiomatic Attribution for Deep Networks
- Understanding Neural Networks Through Deep Visualization
- Understanding Black-box Predictions via Influence Functions
- SAGA: A Fast Incremental Gradient Method With Support for Non-Strongly Convex Composite Objectives
- SmoothGrad: removing noise by adding noise
- Learning both Weights and Connections for Efficient Neural Networks
- Sanity Checks for Saliency Maps
- Actionable Recourse in Linear Classification
- Understanding the Role of Individual Units in a Deep Neural Network
- The Hidden Assumptions Behind Counterfactual Explanations and Principal Reasons
- Explaining Classifiers with Causal Concept Effect (CaCE)
- Adversarial Robustness as a Prior for Learned Representations
- NBDT: Neural-Backed Decision Trees
- From ImageNet to Image Classification: Contextualizing Progress on Benchmarks
- Noise or Signal: The Role of Image Backgrounds in Object Recognition
- Towards falsifiable interpretability research
- Optimal mini-batch and step sizes for SAGA
- Debugging Tests for Model Explanations