Deep Learning for Case-Based Reasoning through Prototypes: A Neural Network that Explains Its Predictions

Abstract

Deep neural networks are widely used for classification. These deep models often suffer from a lack of interpretability -- they are particularly difficult to understand because of their non-linear nature. As a result, neural networks are often treated as "black box" models, and in the past, have been trained purely to optimize the accuracy of predictions. In this work, we create a novel network architecture for deep learning that naturally explains its own reasoning for each prediction. This architecture contains an autoencoder and a special prototype layer, where each unit of that layer stores a weight vector that resembles an encoded training input. The encoder of the autoencoder allows us to do comparisons within the latent space, while the decoder allows us to visualize the learned prototypes. The training objective has four terms: an accuracy term, a term that encourages every prototype to be similar to at least one encoded input, a term that encourages every encoded input to be close to at least one prototype, and a term that encourages faithful reconstruction by the autoencoder. The distances computed in the prototype layer are used as part of the classification process. Since the prototypes are learned during training, the learned network naturally comes with explanations for each prediction, and the explanations are loyal to what the network actually computes.

The first two authors contributed equally, 8 pages, accepted in AAAI 2018

Cited by in corpus (60)

• This Looks Like That: Deep Learning for Interpretable Image Recognition
• Towards Robust Interpretability with Self-Explaining Neural Networks
• Deep k-Nearest Neighbors: Towards Confident, Interpretable and Robust Deep Learning
• Concept Whitening for Interpretable Image Recognition
• Learning Certifiably Optimal Rule Lists for Categorical Data
• A Spatio-Temporal Spot-Forecasting Framework for Urban Traffic Prediction
• This Looks Like That, Because ... Explaining Prototypes for Interpretable Image Recognition
• Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead
• Tree Space Prototypes: Another Look at Making Tree Ensembles Interpretable
• Neural Network Attributions: A Causal Perspective
• A Survey on Neural Network Interpretability
• Interpretability Beyond Classification Output: Semantic Bottleneck Networks
• Privacy Preservation in Federated Learning: An insightful survey from the GDPR Perspective
• Interpreting Deep Learning Models in Natural Language Processing: A Review
• Independent Prototype Propagation for Zero-Shot Compositionality
• Explaining Deep Classification of Time-Series Data with Learned Prototypes
• Embedding Deep Networks into Visual Explanations
• Splitting Steepest Descent for Growing Neural Architectures
• This Looks Like That... Does it? Shortcomings of Latent Space Prototype Interpretability in Deep Networks
• Distribution Density, Tails, and Outliers in Machine Learning: Metrics and Applications
• Synthetic Benchmarks for Scientific Research in Explainable Machine Learning
• ProtoAttend: Attention-Based Prototypical Learning
• Generative causal explanations of black-box classifiers
• Interpretable Anomaly Detection with DIFFI: Depth-based Isolation Forest Feature Importance
• Personalized explanation in machine learning: A conceptualization
• Melody: Generating and Visualizing Machine Learning Model Summary to Understand Data and Classifiers Together
• The Twin-System Approach as One Generic Solution for XAI: An Overview of ANN-CBR Twins for Explaining Deep Learning
• SUBPLEX: Towards a Better Understanding of Black Box Model Explanations at the Subpopulation Level
• Attribute Prototype Network for Zero-Shot Learning
• XProtoNet: Diagnosis in Chest Radiography with Global and Local Explanations
• To Trust or Not to Trust a Regressor: Estimating and Explaining Trustworthiness of Regression Predictions
• Interpreting Neural Networks Using Flip Points
• MonoNet: Towards Interpretable Models by Learning Monotonic Features
• Learning Optimal Decision Trees from Large Datasets
• Knowledge-based XAI through CBR: There is more to explanations than models can tell
• GrCAN: Gradient Boost Convolutional Autoencoder with Neural Decision Forest
• Mediation Challenges and Socio-Technical Gaps for Explainable Deep Learning Applications
• Analyzing the Interpretability Robustness of Self-Explaining Models
• Concept Learners for Few-Shot Learning
• Towards Privacy-preserving Explanations in Medical Image Analysis
• YASENN: Explaining Neural Networks via Partitioning Activation Sequences
• Toward a Unified Framework for Debugging Concept-based Models
• Global Semantic Description of Objects based on Prototype Theory
• Computing Optimal Decision Sets with SAT
• Reasons, Values, Stakeholders: A Philosophical Framework for Explainable Artificial Intelligence
• Introducing the structural bases of typicality effects in deep learning
• Using Wavelets to Analyze Similarities in Image-Classification Datasets
• Toward Scalable and Unified Example-based Explanation and Outlier Detection
• On Explaining Random Forests with SAT
• Towards Fully Interpretable Deep Neural Networks: Are We There Yet?
• Explainability-aided Domain Generalization for Image Classification
• Not All Features Are Equal: Feature Leveling Deep Neural Networks for Better Interpretation
• Training Machine Learning Models by Regularizing their Explanations
• Prediction of Hereditary Cancers Using Neural Networks
• Confusion-based rank similarity filters for computationally-efficient machine learning on high dimensional data
• Inductive Granger Causal Modeling for Multivariate Time Series
• How to Explain Neural Networks: an Approximation Perspective
• It's FLAN time! Summing feature-wise latent representations for interpretability
• Learning to Predict with Supporting Evidence: Applications to Clinical Risk Prediction
• CACTUS: Detecting and Resolving Conflicts in Objective Functions