/
mutual_info_score.py
79 lines (60 loc) · 2.55 KB
/
mutual_info_score.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
# Copyright The Lightning team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import torch
from torch import Tensor, tensor
from torchmetrics.functional.clustering.utils import calculate_contingency_matrix, check_cluster_labels
def _mutual_info_score_update(preds: Tensor, target: Tensor) -> Tensor:
"""Update and return variables required to compute the mutual information score.
Args:
preds: predicted class labels
target: ground truth class labels
Returns:
contingency: contingency matrix
"""
check_cluster_labels(preds, target)
return calculate_contingency_matrix(preds, target)
def _mutual_info_score_compute(contingency: Tensor) -> Tensor:
"""Compute the mutual information score based on the contingency matrix.
Args:
contingency: contingency matrix
Returns:
mutual_info: mutual information score
"""
n = contingency.sum()
u = contingency.sum(dim=1)
v = contingency.sum(dim=0)
# Check if preds or target labels only have one cluster
if u.size() == 1 or v.size() == 1:
return tensor(0.0)
# Find indices of nonzero values in U and V
nzu, nzv = torch.nonzero(contingency, as_tuple=True)
contingency = contingency[nzu, nzv]
# Calculate MI using entries corresponding to nonzero contingency matrix entries
log_outer = torch.log(u[nzu]) + torch.log(v[nzv])
mutual_info = contingency / n * (torch.log(n) + torch.log(contingency) - log_outer)
return mutual_info.sum()
def mutual_info_score(preds: Tensor, target: Tensor) -> Tensor:
"""Compute mutual information between two clusterings.
Args:
preds: predicted cluster labels
target: ground truth cluster labels
Example:
>>> from torchmetrics.functional.clustering import mutual_info_score
>>> target = torch.tensor([0, 3, 2, 2, 1])
>>> preds = torch.tensor([1, 3, 2, 0, 1])
>>> mutual_info_score(preds, target)
tensor(1.0549)
"""
contingency = _mutual_info_score_update(preds, target)
return _mutual_info_score_compute(contingency)