%0 Journal Article %1 tsne %A van der Maaten, Laurens %A Hinton, Geoffrey E. %D 2008 %J Journal of Machine Learning Research %K dimensionality_reduction visualization %P 2579-2605 %T Visualizing High-Dimensional Data Using t-SNE %V 9 %X We present a new technique called “t-SNE” that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces signi&$\#$64257;cantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to in&$\#$64258;uence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are signi&$\#$64257;cantly better than those produced by the other techniques on almost all of the datasets.

@article{tsne, abstract = {We present a new technique called {\textquotedblleft}t-SNE{\textquotedblright} that visualizes high-dimensional data by giving each datapoint a location in a two or three-dimensional map. The technique is a variation of Stochastic Neighbor Embedding (Hinton and Roweis, 2002) that is much easier to optimize, and produces signi\&$\#$64257;cantly better visualizations by reducing the tendency to crowd points together in the center of the map. t-SNE is better than existing techniques at creating a single map that reveals structure at many different scales. This is particularly important for high-dimensional data that lie on several different, but related, low-dimensional manifolds, such as images of objects from multiple classes seen from multiple viewpoints. For visualizing the structure of very large datasets, we show how t-SNE can use random walks on neighborhood graphs to allow the implicit structure of all of the data to in\&$\#$64258;uence the way in which a subset of the data is displayed. We illustrate the performance of t-SNE on a wide variety of datasets and compare it with many other non-parametric visualization techniques, including Sammon mapping, Isomap, and Locally Linear Embedding. The visualizations produced by t-SNE are signi\&$\#$64257;cantly better than those produced by the other techniques on almost all of the datasets.}, added-at = {2015-03-24T16:21:30.000+0100}, attachments = {http://prlab.tudelft.nl/sites/default/files/vandermaaten08a.pdf}, author = {van der Maaten, Laurens and Hinton, Geoffrey E.}, biburl = {https://www.bibsonomy.org/bibtex/25fde64f4b6ca95060a329b980a54013f/saos}, interhash = {d3658497eff9fea21e66bb43320a822a}, intrahash = {5fde64f4b6ca95060a329b980a54013f}, journal = {Journal of Machine Learning Research}, keywords = {dimensionality_reduction visualization}, pages = {2579-2605}, timestamp = {2015-04-29T17:59:14.000+0200}, title = {Visualizing High-Dimensional Data Using t-SNE}, volume = 9, year = 2008 }