Manifold.TSNE

T-distributed Stochastic Neighbor Embedding for visualizing high dimensional data.

TSNE models pairwise similarities both in the original space and in the embedding. It iteratively updates the embedding using gradient descent to minimize the Kullback–Leibler divergence between these two distributions.

constructor(options: TSNEOptions = {})

Options

• nComponents (number, default 2): embedding dimensions
• perplexity (number, default 30)
• learningRate (number, default 200)
• nIter (number, default 250)

Methods

• fit(X: number[][]): void
• fitTransform(X: number[][]): number[][]
• getEmbedding(): number[][]

Example

const tsne = new TSNE({ perplexity: 20 });
const Y = tsne.fitTransform(X);