Scikit-Learn Manifold Methods
The following Scikit-Learn module types are non-linear dimensionality reduction methods.
Modules
ClassicalMDS
Scikit-Learn module documentation: MDS
ClassicalMDS is implemented using sklearn’s MDS with metric=True (Principal Coordinate Analysis / Classical Multidimensional Scaling).
Arguments
| Description | Zero2Neuro Argument | Possible Values | Default Value | Scikit-Learn Parameter |
|---|---|---|---|---|
| Number of dimensions for the embedding | --skl_n_components | Integer ≥ 1 | None | n_components |
| Number of initializations | --skl_n_init | Integer ≥ 1 | None | n_init |
| Maximum number of iterations | --skl_max_iter | Integer ≥ 1 | None | max_iter |
| Convergence tolerance | --skl_tol | Float ≥ 0 | None | eps |
| Random seed for reproducibility | --skl_random_state | Integer | None | random_state |
Isomap
Scikit-Learn module documentation: Isomap
Arguments
| Description | Zero2Neuro Argument | Possible Values | Default Value | Scikit-Learn Parameter |
|---|---|---|---|---|
| Number of CPU cores to use | --cpus_per_task | Integer ≥ 1 | None | n_jobs |
| Number of nearest neighbors | --skl_n_neighbors | Integer ≥ 1 | None | n_neighbors |
| Number of dimensions for the embedding | --skl_n_components | Integer ≥ 1 | None | n_components |
| Distance metric | --skl_metric | String; see sklearn docs | None | metric |
LocallyLinearEmbedding
Scikit-Learn module documentation: LocallyLinearEmbedding
Arguments
| Description | Zero2Neuro Argument | Possible Values | Default Value | Scikit-Learn Parameter |
|---|---|---|---|---|
| Number of CPU cores to use | --cpus_per_task | Integer ≥ 1 | None | n_jobs |
| Number of nearest neighbors | --skl_n_neighbors | Integer ≥ 1 | None | n_neighbors |
| Number of dimensions for the embedding | --skl_n_components | Integer ≥ 1 | None | n_components |
| Maximum number of iterations | --skl_max_iter | Integer ≥ 1 | None | max_iter |
| Convergence tolerance | --skl_tol | Float ≥ 0 | None | tol |
| LLE algorithm variant | --skl_method | standard | hessian | modified | ltsa | None | method |
| Random seed for reproducibility | --skl_random_state | Integer | None | random_state |
MDS
Scikit-Learn module documentation: MDS
Arguments
| Description | Zero2Neuro Argument | Possible Values | Default Value | Scikit-Learn Parameter |
|---|---|---|---|---|
| Number of CPU cores to use | --cpus_per_task | Integer ≥ 1 | None | n_jobs |
| Number of dimensions for the embedding | --skl_n_components | Integer ≥ 1 | None | n_components |
| Number of initializations | --skl_n_init | Integer ≥ 1 | None | n_init |
| Maximum number of iterations | --skl_max_iter | Integer ≥ 1 | None | max_iter |
| Convergence tolerance | --skl_tol | Float ≥ 0 | None | eps |
| Random seed for reproducibility | --skl_random_state | Integer | None | random_state |
SpectralEmbedding
Scikit-Learn module documentation: SpectralEmbedding
Arguments
| Description | Zero2Neuro Argument | Possible Values | Default Value | Scikit-Learn Parameter |
|---|---|---|---|---|
| Number of CPU cores to use | --cpus_per_task | Integer ≥ 1 | None | n_jobs |
| Number of dimensions for the embedding | --skl_n_components | Integer ≥ 1 | None | n_components |
| Random seed for reproducibility | --skl_random_state | Integer | None | random_state |
TSNE
Scikit-Learn module documentation: TSNE
Arguments
| Description | Zero2Neuro Argument | Possible Values | Default Value | Scikit-Learn Parameter |
|---|---|---|---|---|
| Number of CPU cores to use | --cpus_per_task | Integer ≥ 1 | None | n_jobs |
| Number of dimensions for the embedding | --skl_n_components | Integer ≥ 1 | None | n_components |
| Perplexity for nearest neighbor consideration | --skl_perplexity | Float > 0 | None | perplexity |
| Learning rate for optimization | --skl_shrinkage | Float > 0 | None | learning_rate |
| Maximum number of iterations | --skl_max_iter | Integer ≥ 1 | None | max_iter |
| Random seed for reproducibility | --skl_random_state | Integer | None | random_state |