HOLE: Homological Observation of Latent Embeddings for Neural Network Interpretability

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Athreya, Sudhanva Manjunath, Rosen, Paul
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908940406620160
author Athreya, Sudhanva Manjunath
Rosen, Paul
author_facet Athreya, Sudhanva Manjunath
Rosen, Paul
contents Deep learning models have achieved remarkable success across various domains, yet their learned representations and decision-making processes remain largely opaque and hard to interpret. This work introduces HOLE (Homological Observation of Latent Embeddings), a method for analyzing and interpreting discriminative neural networks through persistent homology. HOLE extracts topological features from intermediate activations and presents them using a suite of visualization techniques, including cluster flow diagrams, blob graphs, and heatmap dendrograms. These tools facilitate the examination of representation structure and quality across layers. We evaluate HOLE using a range of discriminative models, focusing on representation quality, interpretability across layers, and robustness to input perturbations and model compression. The results indicate that topological analysis reveals patterns associated with class separation, feature disentanglement, and model robustness, providing a complementary perspective for understanding and improving deep learning systems.
format Preprint
id arxiv_https___arxiv_org_abs_2512_07988
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle HOLE: Homological Observation of Latent Embeddings for Neural Network Interpretability
Athreya, Sudhanva Manjunath
Rosen, Paul
Machine Learning
Graphics
Human-Computer Interaction
Deep learning models have achieved remarkable success across various domains, yet their learned representations and decision-making processes remain largely opaque and hard to interpret. This work introduces HOLE (Homological Observation of Latent Embeddings), a method for analyzing and interpreting discriminative neural networks through persistent homology. HOLE extracts topological features from intermediate activations and presents them using a suite of visualization techniques, including cluster flow diagrams, blob graphs, and heatmap dendrograms. These tools facilitate the examination of representation structure and quality across layers. We evaluate HOLE using a range of discriminative models, focusing on representation quality, interpretability across layers, and robustness to input perturbations and model compression. The results indicate that topological analysis reveals patterns associated with class separation, feature disentanglement, and model robustness, providing a complementary perspective for understanding and improving deep learning systems.
title HOLE: Homological Observation of Latent Embeddings for Neural Network Interpretability
topic Machine Learning
Graphics
Human-Computer Interaction
url https://arxiv.org/abs/2512.07988