Relating tSNE and UMAP to Classical Dimensionality Reduction
Not on the map. Why?
Abstract (arXiv)
It has become standard to use gradient-based dimensionality reduction (DR) methods like tSNE and UMAP when explaining what AI models have learned. This makes sense: these methods are fast, robust, and have an uncanny ability to find semantic patterns in high-dimensional data without supervision. Despite this, gradient-based DR methods lack the most important quality that an explainability method should possess: themselves being explainable. That is, given a UMAP output, it is currently unclear what one can say about the corresponding input. We work towards closing this question by relating UMAP to classical DR techniques. Specifically, we show that one can fully recover methods like PCA, MDS, and ISOMAP in the modern DR paradigm: by applying attractions and repulsions onto a randomly initialized dataset. We also show that, with a small change, Locally Linear Embeddings (LLE) can indistinguishably reproduce UMAP outputs. This implies that the UMAP effective objective is minimized by this modified version of LLE (and vice versa). Given this, we discuss what must be true of UMAP emebddings and present avenues for future work.
Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 39 of its 81 references that are in the map. First posted on arXiv 20 Jun 2023.
Builds on
39 of 81 references.
References · 39
- 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
- 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,349
- 2018UMAP: Uniform Manifold Approximation and Projection for Dimension Reductioncited by 2,898
- 2020Prototypical Contrastive Learning of Unsupervised Representationscited by 640
- 2022MedCLIP: Contrastive Learning from Unpaired Medical Images and Textcited by 570
- 2022Unified Contrastive Learning in Image-Text-Label Spacecited by 194
- 2020Understanding How Dimension Reduction Tools Work: An Empirical Approach to Deciphering t-SNE, UMAP, TriMAP, and PaCMAP for Data Visualizationcited by 173
- 2021Decoupled Contrastive Learningcited by 143
- 2020Cold-start Active Learning through Self-supervised Language Modelingcited by 128
- 2021Supporting Clustering with Contrastive Learningcited by 90
- 2021Contrastive Learning of Musical Representationscited by 82
- 2016Visualizing Large-scale and High-dimensional Datacited by 79
27 more
- 2021A Broad Study on the Transferability of Visual Representations with Contrastive Learningcited by 78
- 2020Attract, Perturb, and Explore: Learning a Feature Alignment Network for Semi-supervised Domain Adaptationcited by 69
- 2021CoLA: Weakly-Supervised Temporal Action Localization with Snippet Contrastive Learningcited by 69
- 2019TriMap: Large-scale Dimensionality Reduction Using Tripletscited by 60
- 2021Your Classifier can Secretly Suffice Multi-Source Domain Adaptationcited by 42
- 2020Attraction-Repulsion Spectrum in Neighbor Embeddingscited by 41
- 2019When Explainability Meets Adversarial Learning: Detecting Adversarial Examples using SHAP Signaturescited by 40
- 2019Unsupervised Domain Adaptation via Regularized Conditional Alignmentcited by 34
- 2021On UMAP's true loss functioncited by 32
- 2020Feature Alignment and Restoration for Domain Generalization and Adaptationcited by 31
- 2020Domain Generalization via Optimal Transport with Metric Similarity Learningcited by 27
- 2021Minimum-Distortion Embeddingcited by 26
- 2019Half a Percent of Labels is Enough: Efficient Animal Detection in UAV Imagery using Deep CNNs and Active Learningcited by 22
- 2017Sublinear Time Low-Rank Approximation of Positive Semidefinite Matricescited by 18
- 2020Locally Linear Embedding and its Variants: Tutorial and Surveycited by 15
- 2021Unsupervised Hyperbolic Representation Learning via Message Passing Auto-Encoderscited by 15
- 2019Unsupervised and Supervised Principal Component Analysis: Tutorialcited by 12
- 2021Reducing the Covariate Shift by Mirror Samples in Cross Domain Alignmentcited by 10
- 2018Parallel Transport Unfolding: A Connection-based Manifold Learning Approachcited by 9
- 2022Studying the impact of magnitude pruning on contrastive learning methodscited by 7
- 2021Incremental Multi-Target Domain Adaptation for Object Detection with Efficient Domain Transfercited by 6
- 2022Unsupervised Contrastive Domain Adaptation for Semantic Segmentationcited by 6
- 2023ActUp: Analyzing and Consolidating tSNE and UMAPcited by 5
- 2021Domain Adaptation for Sentiment Analysis Using Increased Intraclass Separationcited by 4
- 2021Dynamic Feature Alignment for Semi-supervised Domain Adaptationcited by 4
- 2022Local Explanation of Dimensionality Reductioncited by 4
- 2021Anomaly Detection in IR Images of PV Modules using Supervised Contrastive Learningcited by 3
Built on by
Citing papers per year: 2024: 1; 2025: 2.
3 papers.