Concept Evolution in Deep Learning Training: A Unified Interpretation Framework and Discoveries

TL;DR

ConceptEvo is a unified framework that interprets and tracks the evolution of learned concepts in deep neural networks during training, providing insights into model development and decision-making.

Contribution

It introduces a novel algorithm for creating a shared semantic space and discovering concept evolutions, advancing post-training interpretation to include training dynamics.

Findings

Successfully identifies concept evolutions across models

Concept evolutions are human-understandable and influence predictions

Applicable to various modern and classic DNN architectures

Abstract

We present ConceptEvo, a unified interpretation framework for deep neural networks (DNNs) that reveals the inception and evolution of learned concepts during training. Our work addresses a critical gap in DNN interpretation research, as existing methods primarily focus on post-training interpretation. ConceptEvo introduces two novel technical contributions: (1) an algorithm that generates a unified semantic space, enabling side-by-side comparison of different models during training, and (2) an algorithm that discovers and quantifies important concept evolutions for class predictions. Through a large-scale human evaluation and quantitative experiments, we demonstrate that ConceptEvo successfully identifies concept evolutions across different models, which are not only comprehensible to humans but also crucial for class predictions. ConceptEvo is applicable to both modern DNN architectures, such as ConvNeXt, and classic DNNs, such as VGGs and InceptionV3.