ica#

diffsptk.ICA#: alias of IndependentComponentAnalysis

class diffsptk.IndependentComponentAnalysis(order: int, n_comp: int, *, func: str = 'logcosh', n_iter: int = 100, eps: float = 0.0001, batch_size: int | None = None, seed: int | None = None, verbose: bool = False, device: device | None = None, dtype: dtype | None = None)[source]#

Independent component analysis module. Note that the forward method is not differentiable.

Parameters:

orderint >= 0: The order of the vector, \(M\).
n_compint >= 1: The number of components, \(K\).
func[‘logcosh’, ‘gauss’]: The nonquadratic function used in the approximation of negentropy.
n_iterint >= 1: The number of iterations.
epsfloat >= 0: The convergence threshold.
batch_sizeint >= 1 or None: The batch size.
seedint or None: The random seed.
verbosebool: If True, shows progress bars.
devicetorch.device or None: The device of this module.
dtypetorch.dtype or None: The data type of this module.

References

[1]

A. Hyvarinen and E. Oja, “Independent component analysis: algorithms and applications,” Neural Networks, vol. 13, pp. 411-430, 2000.

forward(x: Tensor | DataLoader) → Tensor[source]#

Perform independent component analysis.

Parameters:

xTensor [shape=(T, M+1)] or DataLoader: The input vectors or a DataLoader that yields the input vectors.

Returns:

WTensor [shape=(K, K)]: The separating matrix.

Examples

>>> import diffsptk
>>> import torch
>>> ica = diffsptk.IndependentComponentAnalysis(order=1, n_comp=2, n_iter=10)
>>> x = torch.tensor([[1.0, 2.0], [3.0, 1.0], [2.0, 5.0], [6.0, 4.0]])
>>> W = ica(x)
>>> W
tensor([[ 0.8143,  0.5805],
        [-0.5805,  0.8143]])
>>> s = ica.transform(x)
>>> s
tensor([[ 0.8966, -0.6918],
        [ 1.0956,  0.7138],
        [-0.9212, -1.2381],
        [-1.0710,  1.2161]])

transform(x: Tensor) → Tensor[source]#

Separate the input vectors into independent components.

Parameters:

xTensor [shape=(…, M+1)]: The input vectors.

Returns:

outTensor [shape=(…, K)]: The estimated independent components.