Skip to content
t-SNE and dynamic t-SNE in theano
Python
Branch: master
Clone or download

Latest commit

Fetching latest commit…
Cannot retrieve the latest commit at this time.

Files

Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
examples
model
LICENSE.txt
README.md
__init__.py

README.md

thesne

This project is intended as a flexible implementation of t-SNE [1] and dynamic t-SNE [2].

The t-SNE cost function is defined symbolically and automatically translated into efficient (CPU or GPU) code using theano. Because the derivatives are also computed automatically, testing alternative cost functions becomes very easy.

Dynamic t-SNE is an adaptation of t-SNE for sequences of time-dependent datasets. It introduces a controllable trade-off between temporal coherence and projection (embedding) reliability. For more details, please see [2].

If your use of this code results in a publication, please cite (at least) the original paper by Laurens van der Maaten [1].

This implementation is not nearly as computationally efficient as some alternatives (e.g., Barnes-Hut t-SNE [3]).

Examples

See the examples directory. Remember to add the directory that contains thesne to the PYTHONPATH.

References

[1] Van der Maaten, Laurens and Hinton, Geoffrey. Visualizing data using t-SNE. Journal of Machine Learning Research, 2008.

[2] Rauber, Paulo E., Falcão, Alexandre X., Telea, Alexandru C. Visualizing Time-Dependent Data Using Dynamic t-SNE. Proc. EuroVis Short Papers, 2016.

[3] Van der Maaten, Laurens. Accelerating t-SNE using Tree-Based Algorithms. Journal of Machine Learning Research, 2014.

You can’t perform that action at this time.