Seongmin Lee CV

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

Full Paper, ACM International Conference on Information and Knowledge Management (CIKM), 2023
Poster Paper, Artificial Intelligence & Human-Computer Interaction Workshop at ICML (AI&HCI), 2023


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.


  title={ConceptEvo: Interpreting Concept Evolution in Deep Learning Training},
  author={Park, Haekyu and Lee, Seongmin and Hoover, Benjamin and Wright, Austin and Shaikh, Omar and Duggal, Rahul and Das, Nilaksh and Hoffman, Judy and Chau, Duen Horng},
  journal={arXiv preprint arXiv:2203.16475},