asteroid.losses.mse module

asteroid.losses.mse.MultiSrcMSE

alias of asteroid.losses.mse.SingleSrcMSE

class asteroid.losses.mse.NoSrcMSE(*args, **kwargs)

Bases: asteroid.losses.mse.SingleSrcMSE, asteroid.utils.deprecation_utils.DeprecationMixin

asteroid.losses.mse.NonPitMSE

alias of asteroid.losses.mse.NoSrcMSE

class asteroid.losses.mse.PairwiseMSE(*args, **kwargs)

Bases: sphinx.ext.autodoc.importer._MockObject

Measure pairwise mean square error on a batch.

Shape:

est_targets (torch.Tensor): Expected shape [batch, nsrc, *].

The batch of target estimates.

targets (torch.Tensor): Expected shape [batch, nsrc, *].

The batch of training targets

Returns:torch.Tensor – with shape [batch, nsrc, nsrc]

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(PairwiseMSE(), pit_from='pairwise')
>>> loss = loss_func(est_targets, targets)

forward(est_targets, targets)

class asteroid.losses.mse.SingleSrcMSE(*args, **kwargs)

Bases: sphinx.ext.autodoc.importer._MockObject

Measure mean square error on a batch. Supports both tensors with and without source axis.

Shape:

est_targets (torch.Tensor): Expected shape [batch, *].

The batch of target estimates.

targets (torch.Tensor): Expected shape [batch, *].

The batch of training targets.

Returns:torch.Tensor – with shape [batch]

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> # singlesrc_mse / multisrc_mse support both 'pw_pt' and 'perm_avg'.
>>> loss_func = PITLossWrapper(singlesrc_mse, pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)

forward(est_targets, targets)

asteroid.losses.mse.multisrc_mse

Measure mean square error on a batch. Supports both tensors with and without source axis.

Shape:

est_targets (torch.Tensor): Expected shape [batch, *].

The batch of target estimates.

targets (torch.Tensor): Expected shape [batch, *].

The batch of training targets.

Returns:torch.Tensor – with shape [batch]

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> # singlesrc_mse / multisrc_mse support both 'pw_pt' and 'perm_avg'.
>>> loss_func = PITLossWrapper(singlesrc_mse, pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)

asteroid.losses.mse.nonpit_mse

Measure mean square error on a batch. Supports both tensors with and without source axis.

Shape:

est_targets (torch.Tensor): Expected shape [batch, *].

The batch of target estimates.

targets (torch.Tensor): Expected shape [batch, *].

The batch of training targets.

Returns:torch.Tensor – with shape [batch]

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> # singlesrc_mse / multisrc_mse support both 'pw_pt' and 'perm_avg'.
>>> loss_func = PITLossWrapper(singlesrc_mse, pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)

asteroid.losses.mse.nosrc_mse

Measure mean square error on a batch. Supports both tensors with and without source axis.

Shape:

est_targets (torch.Tensor): Expected shape [batch, *].

The batch of target estimates.

targets (torch.Tensor): Expected shape [batch, *].

The batch of training targets.

Returns:torch.Tensor – with shape [batch]

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> # singlesrc_mse / multisrc_mse support both 'pw_pt' and 'perm_avg'.
>>> loss_func = PITLossWrapper(singlesrc_mse, pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)

asteroid.losses.mse.pairwise_mse

Measure pairwise mean square error on a batch.

Shape:

est_targets (torch.Tensor): Expected shape [batch, nsrc, *].

The batch of target estimates.

targets (torch.Tensor): Expected shape [batch, nsrc, *].

The batch of training targets

Returns:torch.Tensor – with shape [batch, nsrc, nsrc]

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> loss_func = PITLossWrapper(PairwiseMSE(), pit_from='pairwise')
>>> loss = loss_func(est_targets, targets)

asteroid.losses.mse.singlesrc_mse

Measure mean square error on a batch. Supports both tensors with and without source axis.

Shape:

est_targets (torch.Tensor): Expected shape [batch, *].

The batch of target estimates.

targets (torch.Tensor): Expected shape [batch, *].

The batch of training targets.

Returns:torch.Tensor – with shape [batch]

Examples

>>> import torch
>>> from asteroid.losses import PITLossWrapper
>>> targets = torch.randn(10, 2, 32000)
>>> est_targets = torch.randn(10, 2, 32000)
>>> # singlesrc_mse / multisrc_mse support both 'pw_pt' and 'perm_avg'.
>>> loss_func = PITLossWrapper(singlesrc_mse, pit_from='pw_pt')
>>> loss = loss_func(est_targets, targets)