Schemasdecomposition/tsne/v1

tSNEConfig

Variant

t-Distributed Stochastic Neighbor Embedding. See setup_tSNE.

Raw JSONv1Unknown properties rejected

Properties

k: integerdefault 2≥ 1
Number of components to extract.
initial_dims: integerdefault 50≥ 1
Initial dimensions.
perplexity: numberdefault 30≥ 0
Perplexity.
theta: numberdefault 0.5≥ 0≤ 1
Speed/accuracy trade-off.
check_duplicates: booleandefault true
Check for duplicates.
pca: booleandefault true
Perform an initial PCA step.
partial_pca: booleandefault false
Use truncated PCA (irlba).
max_iter: integerdefault 1000≥ 1
Maximum number of iterations.
verbose: booleandefault false
Print progress.
is_distance: booleandefault false
Treat the input as a distance matrix.
pca_center: booleandefault true
Center before the PCA step.
pca_scale: booleandefault false
Scale before the PCA step.
normalize: booleandefault true
Normalize the input.
stop_lying_iter: integerdefault 250≥ 0
Iteration after which exaggeration stops.
mom_switch_iter: integerdefault 250≥ 0
Iteration at which momentum switches.
momentum: numberdefault 0.5
Initial momentum.
final_momentum: numberdefault 0.8
Momentum used later in optimization.
eta: numberdefault 200
Learning rate.
exaggeration_factor: numberdefault 12
Early-exaggeration factor.
num_threads: integerdefault 1≥ 0
Number of threads (0 = all cores).
Y_init: number[][] | null≥ 1 items
Optional initial Y (embedding) matrix. null = random initialization.

Used by