Methoden vergleichen

	UMAP	t-SNE
Fachgebiet	Maschinelles Lernen	Maschinelles Lernen
Familie	Machine learning	Machine learning
Entstehungsjahr	2018	2008
Urheber	McInnes, L.; Healy, J.; Melville, J.	van der Maaten, L. & Hinton, G.
Typ	Nonlinear manifold-learning dimension reduction	Nonlinear dimensionality reduction (manifold visualization)
Wegweisende Quelle	McInnes, L., Healy, J. & Melville, J. (2018). UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction. arXiv:1802.03426.	van der Maaten, L. & Hinton, G. (2008). Visualizing Data using t-SNE. Journal of Machine Learning Research, 9(86), 2579–2605.
Aliasnamen	UMAP (Uniform Manifold Approximation and Projection), uniform manifold approximation and projection, manifold dimension reduction	t-SNE (Boyut İndirgeme / Görselleştirme), t-distributed stochastic neighbor embedding, tsne
Verwandt	5	3
Zusammenfassung	UMAP (Uniform Manifold Approximation and Projection) is a fast, scalable nonlinear dimension-reduction method grounded in manifold-learning theory, introduced by McInnes, Healy and Melville in 2018. It compresses high-dimensional data into a low-dimensional embedding for visualisation and downstream analysis.	t-SNE (t-Distributed Stochastic Neighbor Embedding) is a nonlinear dimensionality-reduction method introduced by Laurens van der Maaten and Geoffrey Hinton in 2008 that maps high-dimensional data into a 2D or 3D space for visualization. It preserves probabilistic local similarities, so points that are neighbours in the original space stay close together, revealing cluster structure and local neighbourhoods.