UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
By Leland McInnes, John Healy. 2018
12,836 citations recorded by the source corpus.
UMAP (Uniform Manifold Approximation and Projection) is a novel manifold learning technique for dimension reduction. UMAP is constructed from a theoretical framework based in Riemannian geometry and algebraic topology. The result is a practical scalable algorithm that applies to real world data. The UMAP algorithm is competitive with t-SNE for visualization quality, and arguably preserves more of the global structure with superior run time performance. Furthermore, UMAP has no computational restrictions on embedding dimension, making it viable as a general purpose dimension reduction technique for machine learning.
Subfields: Aerospace Engineering, Computational Mechanics, Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Aerospace Engineering
- Mohammad Rostami
- Soheil Kolouri
- D. Angelov
- Maarten Grootendorst
- Dmitry Kobak
- Rahul Satija
- David J. Fleet
- Philipp Berens
In Computational Mechanics
- Mohammad Rostami
- Maarten Grootendorst
- Soheil Kolouri
- D. Angelov
- Marta Ruiz Costa-jussà
- Rahul Satija
- Dmitry Kobak
- Blaz Zupan
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Trevor Darrell
- Jitendra Malik
- Ming-Hsuan Yang
- Xiaogang Wang
- Andrew Zisserman
- Andrea Vedaldi
In Representation & Self-Supervised Learning
- Dmitry Kobak
- Philipp Berens
- Mohammad Rostami
- Leland McInnes
- Marta Ruiz Costa-jussà
- Samuel R. Bowman
- J. Böhm
- Cynthia Rudin
Citations in the corpus
- Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned (2022, methodology): “Then we used UMAP [38] to turn the high-dimensional embeddings into two-dimensional embeddings for visualization.”
- Discovering Language Model Behaviors with Model-Written Evaluations (2022, methodology): “, 2020), and we use UMAP (McInnes et al., 2018) to visualize the vectors in a 2D scatter plot, coloring each point by its label.”
- Rescaling Egocentric Vision: Collection, Pipeline and Challenges for EPIC-KITCHENS-100 (2020, background): “UMAP [96] of feature spaces shows better alignment through UDA baseline.”
- Evaluating the Underlying Gender Bias in Contextualized Word Embeddings (2019, methodology): “applying UMAP (McInnes et al., 2018; McInnes et al., 2018) to the contextualized embeddings.”
- Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges (2021, methodology): “Fig 14 shows some DR results for the Mammoth dataset (The Smithsonian Institute, 2020; Coenen and Pearce, 2019) using t-SNE and UMAP with different sets of reasonable hyperparameters.”
- Understanding How Dimension Reduction Tools Work: An Empirical Approach to Deciphering t-SNE, UMAP, TriMAP, and PaCMAP for Data Visualization (2020, methodology): “, 2016) and the current state-of-the art, UMAP (McInnes et al., 2018), are the most well-known of these variants.”
- TriMap: Large-scale Dimensionality Reduction Using Triplets (2019, methodology): “3 hours while LargeVis takes more than 3 hours and UMAP exceeds the 12 hours time limit.”
- openTSNE: a modular Python library for t-SNE dimensionality reduction and embedding (2019, methodology): “Kobak and Linderman (2021) showed that the initialization of force-directed layout algorithms, such as t-SNE, UMAP (McInnes et al. 2018), and ForceAtlas2 (Jacomy, Venturini, Heymann, and Bastian 2014), largely dictates the global consistency of resulting embeddings.”
- The art of using t-SNE for single-cell transcriptomics (2018, background): “there was a continuous progression from radial glia (cluster 7), to neural progenitors (9), to postmitotic premature neurons (10), to mature excitatory (5) and inhibitory (15) neurons.”
- CPR: Classifier-Projection Regularization for Continual Learning (2020, methodology): “This is because, by using UMAP, we verified the change of feature maps of a model with or without CPR and how the plasticity and stability were improved after applying CPR through ablation study.”
- Generative Continual Concept Learning (2019, methodology): “For visualization purpose, we have used UMAP [31], which reduces the dimensionality of the embedding space to two.”
- Complementary Learning for Overcoming Catastrophic Forgetting Using Experience Replay (2019, methodology): “We have used UMAP [McInnes et al., 2018] to reduce the dimensions for visualization purpose.”
- Visualizing and Measuring the Geometry of BERT (2019, methodology): “The system visualizes these 1,000 context embeddings using UMAP [16], generally showing clear clusters relating to word senses.”
- CLIMATE-FEVER: A Dataset for Verification of Real-World Climate Claims (2020, methodology): “We applied umap [19] to find a two-dimensional embedding of the vectors for graphical visualization as can be seen in fig.”
- A Disentangling Invertible Interpretation Network for Explaining Latent Representations (2020, background): “We additionally show a UMAP [35] of the reversed factor manipulations in Fig.”