LGMLOct 19, 2020

A Framework to Learn with Interpretation

arXiv:2010.09345v435 citations
Originality Incremental advance
AI Analysis

This addresses the problem of interpretability for deep learning practitioners, offering a method to generate interpretable models by design or provide post-hoc interpretations, though it appears incremental as it builds on existing interpretability techniques.

The authors tackled interpretability in deep learning by developing a framework that jointly learns a predictive model and an interpretation model, achieving minimal accuracy loss while providing local and global interpretability through high-level attribute functions.

To tackle interpretability in deep learning, we present a novel framework to jointly learn a predictive model and its associated interpretation model. The interpreter provides both local and global interpretability about the predictive model in terms of human-understandable high level attribute functions, with minimal loss of accuracy. This is achieved by a dedicated architecture and well chosen regularization penalties. We seek for a small-size dictionary of high level attribute functions that take as inputs the outputs of selected hidden layers and whose outputs feed a linear classifier. We impose strong conciseness on the activation of attributes with an entropy-based criterion while enforcing fidelity to both inputs and outputs of the predictive model. A detailed pipeline to visualize the learnt features is also developed. Moreover, besides generating interpretable models by design, our approach can be specialized to provide post-hoc interpretations for a pre-trained neural network. We validate our approach against several state-of-the-art methods on multiple datasets and show its efficacy on both kinds of tasks.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes