Source code for torchdr.spectral_embedding.pca
"""Principal Component Analysis module."""
# Authors: Hugues Van Assel <vanasselhugues@gmail.com>
#
# License: BSD 3-Clause License
from typing import Optional, Union, Any
import numpy as np
import torch
from torchdr.base import DRModule
from torchdr.utils import handle_type, svd_flip
[docs]
class PCA(DRModule):
r"""Principal Component Analysis module.
Parameters
----------
n_components : int, default=2
Number of components to project the input data onto.
device : str, default="auto"
Device on which the computations are performed.
verbose : bool, default=False
Whether to print information during the computations.
random_state : float, default=None
Random seed for reproducibility.
svd_driver : str, optional
Name of the cuSOLVER method to be used for torch.linalg.svd.
This keyword argument only works on CUDA inputs.
Available options are: None, gesvd, gesvdj and gesvda.
Defaults to None.
"""
def __init__(
self,
n_components: int = 2,
device: str = "auto",
verbose: bool = False,
random_state: float = None,
svd_driver: Optional[str] = None,
**kwargs,
):
super().__init__(
n_components=n_components,
device=device,
verbose=verbose,
random_state=random_state,
**kwargs,
)
self.svd_driver = svd_driver
self.mean_ = None
self.components_ = None
def _fit_transform(self, X: torch.Tensor, y: Optional[Any] = None) -> torch.Tensor:
"""Fit the PCA model and apply the dimensionality reduction on X.
Parameters
----------
X : torch.Tensor of shape (n_samples, n_features)
Data on which to fit the PCA model and project onto the components.
y : Optional[Any], default=None
Target values (None for unsupervised transformations).
Returns
-------
embedding_ : torch.Tensor of shape (n_samples, n_components)
Projected data.
"""
self.mean_ = X.mean(0, keepdim=True)
U, S, V = torch.linalg.svd(
X - self.mean_, full_matrices=False, driver=self.svd_driver
)
U, V = svd_flip(U, V) # flip eigenvectors' sign to enforce deterministic output
self.components_ = V[: self.n_components]
self.embedding_ = U[:, : self.n_components] * S[: self.n_components]
return self.embedding_
