You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We present a new technique called “t-SNE” that visualizes
3
+
high-dimensional data by giving each datapoint a location in a two
4
+
or three-dimensional map. The technique is a variation of Stochastic
5
+
Neighbor Embedding {[}@hinton:stochastic{]} that is much easier to
6
+
optimize, and produces significantly better visualizations by
7
+
reducing the tendency to crowd points together in the center of the
8
+
map. t-SNE is better than existing techniques at creating a single
9
+
map that reveals structure at many different scales. This is
10
+
particularly important for high-dimensional data that lie on several
11
+
different, but related, low-dimensional manifolds, such as images of
12
+
objects from multiple classes seen from multiple viewpoints. For
13
+
visualizing the structure of very large data sets, we show how t-SNE
14
+
can use random walks on neighborhood graphs to allow the implicit
15
+
structure of all the data to influence the way in which a subset of
16
+
the data is displayed. We illustrate the performance of t-SNE on a
17
+
wide variety of data sets and compare it with many other
18
+
non-parametric visualization techniques, including Sammon mapping,
19
+
Isomap, and Locally Linear Embedding. The visualization produced by
20
+
t-SNE are significantly better than those produced by other
21
+
techniques on almost all of the data sets.
22
+
authors: Laurens van der Maaten and Geoffrey Hinton
23
+
date: 2008-08-11
24
+
description: >
25
+
This page is a reworking of the original t-SNE article using the Computo template. It aims to help authors submitting to the journal by using some advanced formatting features. We warmly thank the authors of t-SNE and the editor of JMLR for allowing us to use their work to illustrate the Computo spirit.
26
+
draft: false
27
+
journal: Computo
28
+
pdf: ''
29
+
repo: published-paper-tsne
30
+
title: Visualizing Data using t-SNE (mock contributon)
We present a new technique called “t-SNE” that visualizes
35
+
high-dimensional data by giving each datapoint a location in a two
36
+
or three-dimensional map. The technique is a variation of Stochastic
37
+
Neighbor Embeddi{[}@hinton:stochastic{]} that is much easier to
38
+
optimize, and produces significantly better visualizations by
39
+
reducing the tendency to crowd points together in the center of the
40
+
map. t-SNE is better than existing techniques at creating a single
41
+
map that reveals structure at many different scales. This is
42
+
particularly important for high-dimensional data that lie on several
43
+
different, but related, low-dimensional manifolds, such as images of
44
+
objects from multiple classes seen from multiple viewpoints. For
45
+
visualizing the structure of very large data sets, we show how t-SNE
46
+
can use random walks on neighborhood graphs to allow the implicit
47
+
structure of all the data to influence the way in which a subset of
48
+
the data is displayed. We illustrate the performance of t-SNE on a
49
+
wide variety of data sets and compare it with many other
50
+
non-parametric visualization techniques, including Sammon mapping,
51
+
Isomap, and Locally Linear Embedding. The visualization produced by
52
+
t-SNE are significantly better than those produced by other
53
+
techniques on almost all of the data sets.
54
+
authors: Laurens van der Maaten and Geoffrey Hinton
55
+
date: 2008-08-11
56
+
description: >
57
+
This page is a reworking of the original t-SNE article using the Computo template. It aims to help authors submitting to the journal by using some advanced formatting features. We warmly thank the authors of t-SNE and the editor of JMLR for allowing us to use their work to illustrate the Computo spirit.
58
+
draft: false
59
+
journal: Computo
60
+
pdf: ''
61
+
repo: published-paper-tsne-R
62
+
title: Visualizing Data using t-SNE (mock contributon)
0 commit comments