@File : _knn.py
@Time : 2025/03/28 11:43:45
@Author : Alejandro Marrero
@Version : 1.0
@Contact : amarrerd@ull.edu.es
@License : (C)Copyright 2025, Alejandro Marrero
@Desc : None
sparseness(instances, archive, k=15)
Computes the sparseness of the instances in the population.
| Parameters: |
-
instances
(Sequence[Instance])
–
Sequence of instances to compute the sparseness.
-
archive
(Sequence[Instance])
–
Archive of instances to compute the sparseness.
-
k
(int, default:
15
)
–
Number of neighbours to use in KNN. Defaults to 15.
|
| Returns: |
-
Sequence[Instance]
–
Sequence[Instance]: Sequence of instances with the sparseness value assigned to each instance.
|
Source code in digneapy/_core/_knn.py
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95 | def sparseness(
instances: Sequence[Instance], archive: Sequence[Instance], k: int = 15
) -> Sequence[Instance]:
"""Computes the sparseness of the instances in the population.
Args:
instances (Sequence[Instance]): Sequence of instances to compute the sparseness.
archive (Sequence[Instance]): Archive of instances to compute the sparseness.
k (int, optional): Number of neighbours to use in KNN. Defaults to 15.
Returns:
Sequence[Instance]: Sequence of instances with the sparseness value assigned to each instance.
"""
num_instances = len(instances)
_instance_desc = np.array([instance.descriptor for instance in instances])
_archive_desc = np.array([instance.descriptor for instance in archive])
combined = (
_instance_desc
if len(archive) == 0
else np.vstack(
[
_instance_desc,
_archive_desc,
]
)
)
result = np.zeros(num_instances)
for i in range(num_instances):
mask = np.ones(num_instances, bool)
mask[i] = False
differences = combined[i] - combined[np.nonzero(mask)]
distances = np.linalg.norm(differences, axis=1)
_neighbors = np.partition(distances, k + 1)[1 : k + 1]
result[i] = np.sum(_neighbors) / k
instances[i].s = result[i]
return result
|
sparseness_only_values(instances, archive, k=15)
Computes the sparseness of the instances in the population.
| Parameters: |
-
instances
(Sequence[Instance])
–
Sequence of instances to compute the sparseness.
-
archive
(Sequence[Instance])
–
Archive of instances to compute the sparseness.
-
k
(int, default:
15
)
–
Number of neighbours to use in KNN. Defaults to 15.
|
| Returns: |
-
ndarray
–
np.ndarray: Numpy array with the sparseness value assigned to each instance.
|
Source code in digneapy/_core/_knn.py
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 | def sparseness_only_values(
instances: Sequence[Instance], archive: Sequence[Instance], k: int = 15
) -> np.ndarray:
"""Computes the sparseness of the instances in the population.
Args:
instances (Sequence[Instance]): Sequence of instances to compute the sparseness.
archive (Sequence[Instance]): Archive of instances to compute the sparseness.
k (int, optional): Number of neighbours to use in KNN. Defaults to 15.
Returns:
np.ndarray: Numpy array with the sparseness value assigned to each instance.
"""
_instance_desc = np.array([instance.descriptor for instance in instances])
_archive_desc = np.array([instance.descriptor for instance in archive])
combined = (
_instance_desc
if len(archive) == 0
else np.vstack(
[
_instance_desc,
_archive_desc,
]
)
)
dist = (
(_instance_desc**2).sum(-1)[:, None]
+ (combined**2).sum(-1)[None, :]
- 2 * _instance_desc @ combined.T
)
dist = np.nan_to_num(dist, nan=np.inf)
# clipping necessary - numerical approx make some distancies negative
dist = np.sqrt(np.clip(dist, min=0.0))
_neighbors = np.partition(dist, k + 1, axis=1)[:, 1 : k + 1]
s_ = np.sum(_neighbors, axis=1) / k
return s_
|