Building and Interpreting Deep Similarity Models

Abstract

Many learning algorithms such as kernel machines, nearest neighbors, clustering, or anomaly detection, are based on the concept of 'distance' or 'similarity'. Before similarities are used for training an actual machine learning model, we would like to verify that they are bound to meaningful patterns in the data. In this paper, we propose to make similarities interpretable by augmenting them with an explanation in terms of input features. We develop BiLRP, a scalable and theoretically founded method to systematically decompose similarity scores on pairs of input features. Our method can be expressed as a composition of LRP explanations, which were shown in previous works to scale to highly nonlinear functions. Through an extensive set of experiments, we demonstrate that BiLRP robustly explains complex similarity models, e.g. built on VGG-16 deep neural network features. Additionally, we apply our method to an open problem in digital humanities: detailed assessment of similarity between historical documents such as astronomical tables. Here again, BiLRP provides insight and brings verifiability into a highly engineered and problem-specific similarity model.

arXiv comment12 pages, 10 figures

References in corpus (18)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Distributed Representations of Words and Phrases and their Compositionality
• A Survey on Deep Learning in Medical Image Analysis
• Show, Attend and Tell: Neural Image Caption Generation with Visual Attention
• Axiomatic Attribution for Deep Networks
• Methods for Interpreting and Understanding Deep Neural Networks
• Explaining NonLinear Classification Decisions with Deep Taylor Decomposition
• Machine Learning of Accurate Energy-Conserving Molecular Force Fields
• Unmasking Clever Hans Predictors and Assessing What Machines Really Learn
• SmoothGrad: removing noise by adding noise
• How to Explain Individual Classification Decisions
• Towards Exact Molecular Dynamics Simulations with Machine-Learned Force Fields
• Not Just a Black Box: Learning Important Features Through Propagating Activation Differences
• Visualizing Deep Neural Network Decisions: Prediction Difference Analysis
• From Clustering to Cluster Explanations via Neural Networks
• Towards Explaining Anomalies: A Deep Taylor Decomposition of One-Class Models
• Explaining Explanations: Axiomatic Feature Interactions for Deep Networks
• Towards a Unified Evaluation of Explanation Methods without Ground Truth

Cited by in corpus (17)

• A Survey on Explainable Artificial Intelligence (XAI): Towards Medical XAI
• Explaining Deep Neural Networks and Beyond: A Review of Methods and Applications
• Higher-Order Explanations of Graph Neural Networks via Relevant Walks
• From Clustering to Cluster Explanations via Neural Networks
• Toward Explainable AI for Regression Models
• The Clever Hans Effect in Unsupervised Learning
• Disentangled Explanations of Neural Network Predictions by Finding Relevant Subspaces
• Analyzing Atomic Interactions in Molecules as Learned by Neural Networks
• Explaining Predictive Uncertainty by Exposing Second-Order Effects
• PredDiff: Explanations and Interactions from Conditional Expectations
• Towards Symbolic XAI -- Explanation Through Human Understandable Logical Relationships Between Features
• Insightful analysis of historical sources at scales beyond human capabilities using unsupervised Machine Learning and XAI
• Preemptively Pruning Clever-Hans Strategies in Deep Neural Networks
• Towards Robust Explanations for Deep Neural Networks
• Toward Scalable and Unified Example-based Explanation and Outlier Detection
• Evolution and Transformation of Scientific Knowledge over the Sphaera Corpus: A Network Study
• It's FLAN time! Summing feature-wise latent representations for interpretability