pytorch_extra_mhirano.experimental.variance_decomposition.variance_decomposition¶
- pytorch_extra_mhirano.experimental.variance_decomposition.variance_decomposition(inputs: torch.Tensor, targets: torch.Tensor, rcond: Optional[float] = None, zero_intercept: bool = False) Tuple[torch.Tensor, torch.Tensor, torch.Tensor][source]¶
- Parameters
inputs (torch.Tensor) – inputs data. B (batch size) x D (Dimension) or B (batch size) x L (input data length) x D (Dimension)
targets (torch.Tensor, optional) – target data. Usually, teaching data. B x 1. For training, this is required.
rcond (float, optional) – See https://pytorch.org/docs/stable/generated/torch.linalg.lstsq.html
zero_intercept (bool, optional) – if True, set intercept to 0.
- Returns
residual of variance decomposition intercept (torch.Tensor): 1 Dim. Zero when zero_intercept is True coefficients (torch.Tensor): D or L x D. Coefficient for each factors.
- Return type
residual (torch.Tensor)