Skip to main content

TSNE

#include <Skigen/Manifold>

template <typename Scalar = double>
class Skigen::TSNE(n_components=2, perplexity=30, learning_rate=200, n_iter=1000, random_state=std::nullopt)

t-Distributed Stochastic Neighbor Embedding (t-SNE).

Nonlinear dimensionality reduction that embeds high-dimensional data in a low-dimensional space by modelling pairwise similarities as joint probabilities with Gaussian kernels in the input space and a Student-t kernel in the embedding space.

Mirrors sklearn.manifold.TSNE.

Parameters:

  • n_components : int, default=2 Dimension of the embedded space (default 2).

  • perplexity : Scalar, default=30 Related to the number of nearest neighbors (default 30.0).

  • learning_rate : Scalar, default=200 Gradient descent step size (default 200.0).

  • n_iter : int, default=1000 Maximum number of iterations (default 1000).

  • random_state : std::optional< uint64_t >, default=std::nullopt Optional RNG seed for reproducibility.

Attributes:

  • embedding : MatrixType Embedding vectors of shape (n_samples, n_components).

  • kl_divergence : Scalar Final KL divergence between P and Q.

Methods​

SKIGEN_PARAMS()​

Fit the t-SNE model: compute an embedding of X.

Parameters:

  • X Input data of shape (n_samples, n_features).

Returns:

  • result Reference to the fitted transformer (*this).

transform(X)​

Return the stored embedding (same as fit result).

Because t-SNE has no parametric mapping, transform simply returns the embedding computed during fit. X is expected to be the same data that was used for fitting.

Parameters:

  • X : MatrixType Data matrix (must match the training data shape).

Returns:

  • result : MatrixType The embedding of shape (n_samples, n_components).

Example​

Plotting​

The figure below is rendered from a registered SkigenPlot-enabled example during the documentation build.

Source example: examples/manifold/tsne.cpp

t-SNE embeddingt-SNE embedding 
Eigen::VectorXi labels(n);
for (int c = 0; c < 3; ++c)
    for (int i = 0; i < n_per; ++i) labels(c * n_per + i) = c;

Skigen::Plot::Figure fig;
fig.title("t-SNE embedding")
   .caption("Three 4-D Gaussian clusters embedded into 2-D by exact Skigen::TSNE")
   .xlabel("t-SNE 1")
   .ylabel("t-SNE 2")
   .scatter(Y, labels);
return argc > 1 ? (fig.saveThemed(argv[1]) ? 0 : 1) : fig.show();