BHTsne

This package is a Julia interface to Van der Maaten's Barnes-Hut C++ implementation of t-Distributed Stochastic Neighbor Embedding. There are also other wrappers available in Python, Matlab, R and Torch.

Installation

Pkg.clone("https://github.com/zhmz90/BHTsne.jl.git")

Usage

	using BHTsne
	using RDatasets	
	using Gadfly
	
	iris = dataset("datasets", "iris") 	
	samples = convert(Array, iris[:,1:4])
	labels  = convert(Array, iris[:,5])
	
	results = bh_tsne(samples, perplexity=30, verbose=true)
	
	p = plot(x=results[:,1], y=results[:,2], color=labels)
	draw(PNG("tsne_of_iris.png", 8inch, 6inch), p)

APIs

	function bh_tsne(samples; no_dims=2, initial_dims=50, perplexity=50, theta=0.5, randseed=-1, verbose=false)

References

• Laurens van der Maaten
• Tips working with t-sne