Alexander Engelsberger 7c30ffe2c7 Automatic Formatting.
2021-04-23 17:25:23 +02:00

266 lines
9.2 KiB
Python

"""ProtoTorch distance functions."""
import numpy as np
import torch
from prototorch.functions.helper import (_check_shapes, _int_and_mixed_shape,
equal_int_shape)
def squared_euclidean_distance(x, y):
r"""Compute the squared Euclidean distance between :math:`\bm x` and :math:`\bm y`.
Compute :math:`{\langle \bm x - \bm y \rangle}_2`
:param `torch.tensor` x: Two dimensional vector
:param `torch.tensor` y: Two dimensional vector
**Alias:**
``prototorch.functions.distances.sed``
"""
expanded_x = x.unsqueeze(dim=1)
batchwise_difference = y - expanded_x
differences_raised = torch.pow(batchwise_difference, 2)
distances = torch.sum(differences_raised, axis=2)
return distances
def euclidean_distance(x, y):
r"""Compute the Euclidean distance between :math:`x` and :math:`y`.
Compute :math:`\sqrt{{\langle \bm x - \bm y \rangle}_2}`
:param `torch.tensor` x: Input Tensor of shape :math:`X \times N`
:param `torch.tensor` y: Input Tensor of shape :math:`Y \times N`
:returns: Distance Tensor of shape :math:`X \times Y`
:rtype: `torch.tensor`
"""
distances_raised = squared_euclidean_distance(x, y)
distances = torch.sqrt(distances_raised)
return distances
def euclidean_distance_v2(x, y):
diff = y - x.unsqueeze(1)
pairwise_distances = (diff @ diff.permute((0, 2, 1))).sqrt()
# Passing `dim1=-2` and `dim2=-1` to `diagonal()` takes the
# batch diagonal. See:
# https://pytorch.org/docs/stable/generated/torch.diagonal.html
distances = torch.diagonal(pairwise_distances, dim1=-2, dim2=-1)
# print(f"{diff.shape=}") # (nx, ny, ndim)
# print(f"{pairwise_distances.shape=}") # (nx, ny, ny)
# print(f"{distances.shape=}") # (nx, ny)
return distances
def lpnorm_distance(x, y, p):
r"""Calculate the lp-norm between :math:`\bm x` and :math:`\bm y`.
Also known as Minkowski distance.
Compute :math:`{\| \bm x - \bm y \|}_p`.
Calls ``torch.cdist``
:param `torch.tensor` x: Two dimensional vector
:param `torch.tensor` y: Two dimensional vector
:param p: p parameter of the lp norm
"""
distances = torch.cdist(x, y, p=p)
return distances
def omega_distance(x, y, omega):
r"""Omega distance.
Compute :math:`{\| \Omega \bm x - \Omega \bm y \|}_p`
:param `torch.tensor` x: Two dimensional vector
:param `torch.tensor` y: Two dimensional vector
:param `torch.tensor` omega: Two dimensional matrix
"""
projected_x = x @ omega
projected_y = y @ omega
distances = squared_euclidean_distance(projected_x, projected_y)
return distances
def lomega_distance(x, y, omegas):
r"""Localized Omega distance.
Compute :math:`{\| \Omega_k \bm x - \Omega_k \bm y_k \|}_p`
:param `torch.tensor` x: Two dimensional vector
:param `torch.tensor` y: Two dimensional vector
:param `torch.tensor` omegas: Three dimensional matrix
"""
projected_x = x @ omegas
projected_y = torch.diagonal(y @ omegas).T
expanded_y = torch.unsqueeze(projected_y, dim=1)
batchwise_difference = expanded_y - projected_x
differences_squared = batchwise_difference**2
distances = torch.sum(differences_squared, dim=2)
distances = distances.permute(1, 0)
return distances
def euclidean_distance_matrix(x, y, squared=False, epsilon=1e-10):
r"""Computes an euclidean distances matrix given two distinct vectors.
last dimension must be the vector dimension!
compute the distance via the identity of the dot product. This avoids the memory overhead due to the subtraction!
- ``x.shape = (number_of_x_vectors, vector_dim)``
- ``y.shape = (number_of_y_vectors, vector_dim)``
output: matrix of distances (number_of_x_vectors, number_of_y_vectors)
"""
for tensor in [x, y]:
if tensor.ndim != 2:
raise ValueError(
"The tensor dimension must be two. You provide: tensor.ndim=" +
str(tensor.ndim) + ".")
if not equal_int_shape([tuple(x.shape)[1]], [tuple(y.shape)[1]]):
raise ValueError(
"The vector shape must be equivalent in both tensors. You provide: tuple(y.shape)[1]="
+ str(tuple(x.shape)[1]) + " and tuple(y.shape)(y)[1]=" +
str(tuple(y.shape)[1]) + ".")
y = torch.transpose(y)
diss = (torch.sum(x**2, axis=1, keepdims=True) - 2 * torch.dot(x, y) +
torch.sum(y**2, axis=0, keepdims=True))
if not squared:
if epsilon == 0:
diss = torch.sqrt(diss)
else:
diss = torch.sqrt(torch.max(diss, epsilon))
return diss
def tangent_distance(signals, protos, subspaces, squared=False, epsilon=1e-10):
r"""Tangent distances based on the tensorflow implementation of Sascha Saralajews
For more info about Tangen distances see
DOI:10.1109/IJCNN.2016.7727534.
The subspaces is always assumed as transposed and must be orthogonal!
For local non sparse signals subspaces must be provided!
- shape(signals): batch x proto_number x channels x dim1 x dim2 x ... x dimN
- shape(protos): proto_number x dim1 x dim2 x ... x dimN
- shape(subspaces): (optional [proto_number]) x prod(dim1 * dim2 * ... * dimN) x prod(projected_atom_shape)
subspace should be orthogonalized
Pytorch implementation of Sascha Saralajew's tensorflow code.
Translation by Christoph Raab
"""
signal_shape, signal_int_shape = _int_and_mixed_shape(signals)
proto_shape, proto_int_shape = _int_and_mixed_shape(protos)
subspace_int_shape = tuple(subspaces.shape)
# check if the shapes are correct
_check_shapes(signal_int_shape, proto_int_shape)
atom_axes = list(range(3, len(signal_int_shape)))
# for sparse signals, we use the memory efficient implementation
if signal_int_shape[1] == 1:
signals = torch.reshape(signals, [-1, np.prod(signal_shape[3:])])
if len(atom_axes) > 1:
protos = torch.reshape(protos, [proto_shape[0], -1])
if subspaces.ndim == 2:
# clean solution without map if the matrix_scope is global
projectors = torch.eye(subspace_int_shape[-2]) - torch.dot(
subspaces, torch.transpose(subspaces))
projected_signals = torch.dot(signals, projectors)
projected_protos = torch.dot(protos, projectors)
diss = euclidean_distance_matrix(projected_signals,
projected_protos,
squared=squared,
epsilon=epsilon)
diss = torch.reshape(
diss, [signal_shape[0], signal_shape[2], proto_shape[0]])
return torch.permute(diss, [0, 2, 1])
else:
# no solution without map possible --> memory efficient but slow!
projectors = torch.eye(subspace_int_shape[-2]) - torch.bmm(
subspaces,
subspaces) # K.batch_dot(subspaces, subspaces, [2, 2])
projected_protos = (protos @ subspaces
).T # K.batch_dot(projectors, protos, [1, 1]))
def projected_norm(projector):
return torch.sum(torch.dot(signals, projector)**2, axis=1)
diss = (torch.transpose(map(projected_norm, projectors)) -
2 * torch.dot(signals, projected_protos) +
torch.sum(projected_protos**2, axis=0, keepdims=True))
if not squared:
if epsilon == 0:
diss = torch.sqrt(diss)
else:
diss = torch.sqrt(torch.max(diss, epsilon))
diss = torch.reshape(
diss, [signal_shape[0], signal_shape[2], proto_shape[0]])
return torch.permute(diss, [0, 2, 1])
else:
signals = signals.permute([0, 2, 1] + atom_axes)
diff = signals - protos
# global tangent space
if subspaces.ndim == 2:
# Scope Projectors
projectors = subspaces #
# Scope: Tangentspace Projections
diff = torch.reshape(
diff, (signal_shape[0] * signal_shape[2], signal_shape[1], -1))
projected_diff = diff @ projectors
projected_diff = torch.reshape(
projected_diff,
(signal_shape[0], signal_shape[2], signal_shape[1]) +
signal_shape[3:],
)
diss = torch.norm(projected_diff, 2, dim=-1)
return diss.permute([0, 2, 1])
# local tangent spaces
else:
# Scope: Calculate Projectors
projectors = subspaces
# Scope: Tangentspace Projections
diff = torch.reshape(
diff, (signal_shape[0] * signal_shape[2], signal_shape[1], -1))
diff = diff.permute([1, 0, 2])
projected_diff = torch.bmm(diff, projectors)
projected_diff = torch.reshape(
projected_diff,
(signal_shape[1], signal_shape[0], signal_shape[2]) +
signal_shape[3:],
)
diss = torch.norm(projected_diff, 2, dim=-1)
return diss.permute([1, 0, 2]).squeeze(-1)
# Aliases
sed = squared_euclidean_distance