User Guide
User_guide
User Manual:
Open the PDF directly: View PDF 
.
Page Count: 4
USER’SGUIDE FOR T-SNE SOFTWARE
User’s Guide for t-SNE Software
Laurens van der Maaten LVDMAATEN@GMAIL.COM
Pattern Recognition and Bioinformatics Group
Delft University of Technology
Mekelweg 4, 2628 CD Delft, The Netherlands
Geoffrey Hinton HINTON@CS.TORONTO.EDU
Department of Computer Science
University of Toronto
6 King’s College Road, M5S 3G4 Toronto, ON, Canada
Editor:
1. Introduction
In this document, we describe the use of the t-SNE software that is publicly available online from
http://lvdmaaten.github.io/lvdmaaten/tsne. Please note that this document does
not describe the t-SNE technique itself; for more information on how t-SNE works, we refer
to (van der Maaten and Hinton, 2008). The user’s guide assumes that the Matlab environment
is used, however, the discussion in section 3 is also of relevance to the use of the t-SNE software in
programs that are written in, e.g., R, Java, or C++.
This document decribes two implementations of t-SNE that are available online: (1) a simple
Matlab implementation, and (2) a fast binary Barnes-Hut implementation with wrappers in various
languages. Implementations of t-SNE in other languages work very similarly. The use of the Matlab
implementation is described in Section 2. Section 3 covers the use of the fast t-SNE implementation.
2. Simple Matlab implementation
The main purpose of the Matlab implementation of t-SNE is to illustrate how the technique works.
The code contains a lot of comments, making it a useful resource in the study of the technique. Mat-
lab’s poor memory management probably makes the codes less appropriate for application on large
real-world datasets. The code requires a working installation of Matlab. No additional toolboxes
are required.
The syntax of the Matlab implementation of t-SNE is the following:
mappedX =tsne(X,labels,no_dims,init_dims,perplexity)
Herein, Xdenotes the N×Ddata matrix, in which rows correspond to the Ninstances and columns
correspond to the Ddimensions. In case the labels are specified, the code plots the intermediate so-
lution every ten iterations. The labels are only used in the visualization of the intermediate solutions:
t-SNE is an unsupervised dimensionality reduction technique. If an empty matrix ([]) is specified
instead of a labels vector, no intermediate solutions are shown. The dimensionality of the visual-
ization constructed by t-SNE can be specified through no dims (the default value for no dims is 2).
1
VAN DER MAATEN AND HINTON
Before running t-SNE, the Matlab code preprocesses the data using PCA, reducing its dimensional-
ity to init dims dimensions (the default value is 30). The perplexity of the Gaussian distributions
that are employed in the high-dimensional space can be specified through perplexity (the default
value is 30). The function returns a N×no dims matrix mappedX that specifies the coordinates of
the Nlow-dimensional datapoints.
We illustrate the use of the simple Matlab implementation with an example. The following
example requires the file mnist train.mat as available from the t-SNE website. Notice that the
example takes approximately an hour to complete.
% Load data
load ’mnist_train.mat’
ind =randperm(size(train_X, 1));
train_X =train_X(ind(1:5000),:);
train_labels =train_labels(ind(1:5000));
% Set parameters
no_dims = 2;
initial_dims = 50;
perplexity = 30;
% Run t−SNE
mappedX =tsne(train_X, [], no_dims,initial_dims,perplexity);
% Plot results
gscatter(mappedX(:,1), mappedX(:,2), train_labels,’o’);
3. Barnes-Hut Implementation
The fast Barnes-Hut implementation of t-SNE that is available online was implemented in C++.
This implementation uses a different algorithm, which is described in detail by van der Maaten
(2014). To compile the code, please execute the following command in the terminal:
cd /path/to/bhtsne
g++ sptree.cpp tsne.cpp −o bh_tsne −O2
The code comes with a Matlab script is available that illustrates how the fast implementation
of t-SNE can be used. The syntax of the Matlab script (which is called fast tsne.m) is roughly
similar to that of the tsne function. It is given by:
mappedX =fast_tsne(X,no_dims,initial_dims,perplexity,theta)
Most of the parameters are identical to those discussed in Section 2, which is why we do not discuss
them in detail here. The parameter theta specified how coarse the Barnes-Hut approximation is:
setting theta to 0runs the original O(N2)t-SNE algorithm, whereas using higher values runs the
O(Nlog N)with increasingly better constant. The value of theta should be between 0and 1, and
its default value is 0.5.
2
USER’SGUIDE FOR T-SNE SOFTWARE
An example of the use of the fast (landmark) version of t-SNE is given below. This example
may also take up to an twenty minutes to complete.
% Load data
load ’mnist_train.mat’
% Set parameters
no_dims = 2;
initial_dims = 50;
perplexity = 30;
% Run t−SNE
mappedX =tsne(train_X, [], no_dims,initial_dims,perplexity);
% Plot results
gscatter(mappedX(:,1), mappedX(:,2), train_labels,’o’);
The example embeds all 60,000 datapoints in the MNIST training set!
There are wrappers available of the C++ code in Matlab, Python, Torch, and R. If you want to
use another language, it should be fairly easy to code one up yourself. The fast implementation
of t-SNE assumes the data and the parameters to be specified in a file called data.dat. The file
data.dat is a binary file of which the structure is outlined in Table 1. The data matrix Xshould
be specified as a concatenation of the instances (i.e., {instance1,instance2, ..., instanceN}).
After executing the binary, the result of the algorithm is written in a file called result.dat, the
structure of which is outlined in Table 2. The low-dimensional embedding mappedX is returned as
a concatenation of instances.
Value Number of elements Type
N1 int32
D1 int32
theta 1 double
perplexity 1 double
no dims 1 int32
XN×Ddouble
Table 1: Structure of the data.dat file.
Value Number of elements Type
N1 int32
no dims 1 int32
mappedX N×no dims double
landmarks Nint32
costs Ndouble
Table 2: Structure of the result.dat file. The last two variables are there for legacy reasons, and
need never be used.
3
VAN DER MAATEN AND HINTON
References
L.J.P. van der Maaten. Accelerating t-sne using tree-based algorithms. Journal of Machine Learning
Research, 15(Oct):3221–3245, 2014.
L.J.P. van der Maaten and G.E. Hinton. Visualizing Data using t-SNE. Journal of Machine Learning
Research, 9(Nov):2579–2605, 2008.
4
