Towards Interpretable Deep Neural Networks by Leveraging Adversarial Examples

Abstract

Sometimes it is not enough for a DNN to produce an outcome. For example, in applications such as healthcare, users need to understand the rationale of the decisions. Therefore, it is imperative to develop algorithms to learn models with good interpretability (Doshi-Velez 2017). An important factor that leads to the lack of interpretability of DNNs is the ambiguity of neurons, where a neuron may fire for various unrelated concepts. This work aims to increase the interpretability of DNNs on the whole image space by reducing the ambiguity of neurons. In this paper, we make the following contributions: 1) We propose a metric to evaluate the consistency level of neurons in a network quantitatively. 2) We find that the learned features of neurons are ambiguous by leveraging adversarial examples. 3) We propose to improve the consistency of neurons on adversarial example subset by an adversarial training algorithm with a consistent loss.

In AAAI-19 Workshop on Network Interpretability for Deep Learning

References in corpus (17)

• Explaining and Harnessing Adversarial Examples
• Intriguing properties of neural networks
• Deep Residual Learning for Image Recognition
• Caffe: Convolutional Architecture for Fast Feature Embedding
• Towards A Rigorous Science of Interpretable Machine Learning
• Striving for Simplicity: The All Convolutional Net
• Going Deeper with Convolutions
• Rich feature hierarchies for accurate object detection and semantic segmentation
• Visualizing Deep Neural Network Decisions: Prediction Difference Analysis
• Adversarial Machine Learning at Scale
• Learning Deep Features for Discriminative Localization
• DeepFool: a simple and accurate method to fool deep neural networks
• Generating Visual Explanations
• Analyzing the Performance of Multilayer Neural Networks for Object Recognition
• Attentive Explanations: Justifying Decisions and Pointing to the Evidence
• Improving Interpretability of Deep Neural Networks with Semantic Information
• Towards Better Analysis of Deep Convolutional Neural Networks

Cited by in corpus (44)

• Adversarial Examples: Attacks and Defenses for Deep Learning
• Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
• Theory-guided hard constraint projection (HCP): a knowledge-based data-driven scientific machine learning method
• Interpreting and Improving Adversarial Robustness of Deep Neural Networks with Neuron Sensitivity
• Understanding and Enhancing the Transferability of Adversarial Examples
• Interpretation of Neural Networks is Fragile
• EAD: Elastic-Net Attacks to Deep Neural Networks via Adversarial Examples
• Boosting Adversarial Attacks with Momentum
• I-MAD: Interpretable Malware Detector Using Galaxy Transformer
• Learning Black-Box Attackers with Transferable Priors and Query Feedback
• Interpreting Adversarial Examples by Activation Promotion and Suppression
• Explainable Text Classification in Legal Document Review A Case Study of Explainable Predictive Coding
• IG2: Integrated Gradient on Iterative Gradient Path for Feature Attribution
• Transferable Adversarial Attacks for Image and Video Object Detection
• Unsupervised Learning of Neural Networks to Explain Neural Networks
• Learning More Robust Features with Adversarial Training
• ConvNets and ImageNet Beyond Accuracy: Understanding Mistakes and Uncovering Biases
• Sequential Interpretability: Methods, Applications, and Future Direction for Understanding Deep Learning Models in the Context of Sequential Data
• Evaluating the Robustness of Bayesian Neural Networks Against Different Types of Attacks
• FANN-on-MCU: An Open-Source Toolkit for Energy-Efficient Neural Network Inference at the Edge of the Internet of Things
• Bioinformatics and Medicine in the Era of Deep Learning
• Category-wise Attack: Transferable Adversarial Examples for Anchor Free Object Detection
• EX-RAY: Distinguishing Injected Backdoor from Natural Features in Neural Networks by Examining Differential Feature Symmetry
• A Framework for Explainable Text Classification in Legal Document Review
• Adversarial Attacks and Defenses: An Interpretation Perspective
• Interpreting Adversarial Examples with Attributes
• Data-Free Learning of Student Networks
• Adversarial Learning with Margin-based Triplet Embedding Regularization
• DNN2LR: Interpretation-inspired Feature Crossing for Real-world Tabular Data
• Explaining AlphaGo: Interpreting Contextual Effects in Neural Networks
• Investigating Robustness and Interpretability of Link Prediction via Adversarial Modifications
• A Unified Game-Theoretic Interpretation of Adversarial Robustness
• FineFool: Fine Object Contour Attack via Attention
• An Empirical Study on the Relation between Network Interpretability and Adversarial Robustness
• Heat and Blur: An Effective and Fast Defense Against Adversarial Examples
• Zero-shot Adversarial Quantization
• Noise Modulation: Let Your Model Interpret Itself
• Interpretable Few-Shot Learning via Linear Distillation
• Adversarial Ranking Attack and Defense
• Understanding Misclassifications by Attributes
• Prediction of Hereditary Cancers Using Neural Networks
• Adaptive Gradient for Adversarial Perturbations Generation
• Multi-concept adversarial attacks
• Deep Active Learning by Model Interpretability