Globally local and fast explanations of t-SNE-like nonlinear embeddings

Pierre Lambert, Rebecca Marion, Julien Albert, Emmanuel Jean, Sacha Corbugy, Cyril de Bodt

Research output: Contribution in Book/Catalog/Report/Conference proceedingChapter (peer-reviewed)peer-review

Abstract

Nonlinear dimensionality reduction (NLDR) algorithms such as t-SNE are often employed to visually analyze high-dimensional (HD) data sets in the form of low-dimensional (LD) embeddings. Unfortunately, the nonlinearity of the NLDR process prohibits the interpretation of the resulting embeddings in terms of the HD features. State-of-the-art studies propose post-hoc explanation approaches to locally explain the embeddings. However, such tools are typically slow and do not automatically cover the entire LD embedding, instead providing local explanations around one selected data point at a time. This prevents users from quickly gaining insights about the general explainability landscape of the embedding. This paper presents a globally local and fast explanation framework for NLDR embeddings. This framework is fast because it only requires the computation of sparse linear regression models on subsets of the data, without ever reapplying the NLDR algorithm itself. In addition, the framework is globally local in the sense that the entire LD embedding is automatically covered by multiple local explanations. The different interpretable structures in the embedding are directly characterized, making it possible to quantify the importance of the HD features in various regions of the LD embedding. An example use-case is examined, emphasizing the value of the presented framework. Public codes and a software are available at https://github.com/PierreLambert3/glocally_explained.

Original languageEnglish
Title of host publicationCIKM-WS 2022
Subtitle of host publicationProceedings of the CIKM 2022 Workshops
EditorsGeorgios Drakopoulos, Eleanna Kafeza
PublisherCEUR Workshop Proceedings
Publication statusPublished - 2022
Event2022 International Conference on Information and Knowledge Management Workshops, CIKM-WS 2022 - Atlanta, United States
Duration: 17 Oct 202221 Oct 2022

Publication series

NameCEUR Workshop Proceedings
PublisherCEUR-WS
Volume3318
ISSN (Print)1613-0073

Conference

Conference2022 International Conference on Information and Knowledge Management Workshops, CIKM-WS 2022
Country/TerritoryUnited States
CityAtlanta
Period17/10/2221/10/22

Keywords

  • data exploration
  • data visualization
  • dimensionality reduction
  • explainability
  • interactivity
  • interpretability
  • t-SNE

Fingerprint

Dive into the research topics of 'Globally local and fast explanations of t-SNE-like nonlinear embeddings'. Together they form a unique fingerprint.

Cite this