Title: Topological structure of complex predictions

Journal Article · · Nature Machine Intelligence

Abstract Current complex prediction models are the result of fitting deep neural networks, graph convolutional networks or transducers to a set of training data. A key challenge with these models is that they are highly parameterized, which makes describing and interpreting the prediction strategies difficult. We use topological data analysis to transform these complex prediction models into a simplified topological view of the prediction landscape. The result is a map of the predictions that enables inspection of the model results with more specificity than dimensionality-reduction methods such as tSNE and UMAP. The methods scale up to large datasets across different domains. We present a case study of a transformer-based model previously designed to predict expression levels of a piece of DNA in thousands of genomic tracks. When the model is used to study mutations in the BRCA1 gene, our topological analysis shows that it is sensitive to the location of a mutation and the exon structure of BRCA1 in ways that cannot be found with tools based on dimensionality reduction. Moreover, the topological framework offers multiple ways to inspect results, including an error estimate that is more accurate than model uncertainty. Further studies show how these ideas produce useful results in graph-based learning and image classification.

Research Organization:
Purdue Univ., West Lafayette, IN (United States)
Sponsoring Organization:
USDOE; USDOE Office of Science (SC)
Grant/Contract Number:
SC0023162
OSTI ID:
2216985
Journal Information:
Nature Machine Intelligence, Journal Name: Nature Machine Intelligence Journal Issue: 12 Vol. 5; ISSN 2522-5839
Publisher:
Nature Publishing GroupCopyright Statement
Country of Publication:
United Kingdom
Language:
English

References (25)

ImageNet Large Scale Visual Recognition Challenge journal April 2015
Validation and algorithmic audit of a deep learning system for the detection of proximal femoral fractures in patients in the emergency department: a diagnostic accuracy study journal May 2022
Layered Reeb graphs for three-dimensional manifolds in boundary representation journal February 2015
Computational Topology for Data Analysis book February 2022
Dermatologist-level classification of skin cancer with deep neural networks journal January 2017
Dimensionality reduction for visualizing single-cell data using UMAP journal December 2018
Deep learning and process understanding for data-driven Earth system science journal February 2019
Effective gene expression prediction from sequence by integrating long-range interactions journal October 2021
Towards a topological–geometrical theory of group equivariant non-expansive operators for data analysis and machine learning journal September 2019
Benchmarking saliency methods for chest X-ray interpretation journal October 2022
Extracting insights from the shape of complex data using topology journal February 2013
Topology based data analysis identifies a subgroup of breast cancers with a unique mutational profile and excellent survival journal April 2011
Definitions, methods, and applications in interpretable machine learning journal October 2019
ClinVar: improving access to variant interpretations and supporting evidence journal November 2017
Learning Deep Features for Discriminative Localization conference June 2016
Deep Residual Learning for Image Recognition conference June 2016
ChestX-Ray8: Hospital-Scale Chest X-Ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases conference July 2017
TDA-Net: Fusion of Persistent Homology and Deep Learning Features for COVID-19 Detection From Chest X-Ray Images conference November 2021
Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization conference October 2017
TopoAct: Visually Exploring the Shape of Activations in Deep Learning journal January 2021
Geometric deep learning of RNA structure journal August 2021
Multiscale Mapper: Topological Summarization via Codomain Covers conference December 2015
Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: A cross-sectional study journal November 2018
UMAP: Uniform Manifold Approximation and Projection journal September 2018
MengLiuPurdue/Graph-Topological-Data-Analysis: v0.1 software August 2023

Similar Records

Related Subjects