"TSNE" (Machine Learning Method)

Details & Suboptions

  • "TSNE", which stands for t-distributed stochastic neighbor embedding, is a nonlinear non-parametric dimensionality reduction method. The method attempts to learn a low-dimensional representation of the data that preserves the local structure of the data.
  • "TSNE" works for datasets with nonlinear manifolds and is particularly suited for the visualization of high-dimensional datasets; however, it is slow to train for datasets that have a large number of features and large number of examples.
  • The following shows two-dimensional embeddings learned by the "TSNE" method applied to the benchmark datasets Fisher's Irises, MNIST and FashionMNIST:
  • Given the similarity matrix p_(ij) of data points (xi, xj) in the original space, "TSNE" attempts to find the lower-dimensional embeddings , such that similarities q_(ij) in the lower-dimensional space match p_(ij).
  • Similarities in the original space are given by , where corresponds to a neighborhood radius. Similarities in the lower-dimensional space are given by .
  • The lower-dimensional embeddings are computed by minimizing the embedding cost .
  • The parameter is indirectly controlled by a perplexity parameter. A large perplexity results in a small amount of clusters while small perplexity leads to many clusters.
  • The following suboptions can be given:
  • "Perplexity" Automaticradius ϵ
    "LinearPrereduction" Falsewhether to perform a linear reduction before running the t-SNE method

Examples

open allclose all

Basic Examples  (1)

Reduce the dimension of some images using the "TSNE" method:

Visualize the two-dimensional representation of images:

Options  (2)

"Perplexity"  (1)

Load the Fisher Iris dataset from ExampleData:

Generate a reducer function using the "TSNE" method:

Group the examples by their species:

Reduce the dimension of the features:

Visualize the reduced dataset:

Perform the same operation using a different perplexity value:

"LinearPrereduction"  (1)

Load a sample from the "MNIST" dataset:

Reduce the dimension of images using "TSNE":

Find features by performing a linear reduction before running the t-SNE method using the "LinearPrereduction" suboption:

Visualize the obtained features and compare the results:

Applications  (1)

Data Visualization  (1)

Reduce the dimension of some images using the "TSNE" method:

Visualize the two-dimensional representation of images: