|
1
|
1 |
|
import numpy as np |
|
2
|
1 |
|
import numpy.random as rnd |
|
3
|
1 |
|
import pytest |
|
4
|
1 |
|
from libtlda.util import is_pos_def, one_hot, regularize_matrix |
|
5
|
|
|
|
|
6
|
|
|
|
|
7
|
1 |
|
def test_is_pos_def(): |
|
8
|
|
|
"""Check if function returns boolean positive-definiteness.""" |
|
9
|
|
|
# Positive-definite matrix |
|
10
|
1 |
|
A = np.array([[1, 0], [0, 1]]) |
|
11
|
|
|
|
|
12
|
|
|
# Not positive-definite matrix |
|
13
|
1 |
|
B = np.array([[-1, 0], [0, 1]]) |
|
14
|
|
|
|
|
15
|
|
|
# Assert correct positive-definiteness |
|
16
|
1 |
|
assert is_pos_def(A) |
|
17
|
1 |
|
assert not is_pos_def(B) |
|
18
|
|
|
|
|
19
|
|
|
|
|
20
|
1 |
|
def test_one_hot(): |
|
21
|
|
|
"""Check if one_hot returns correct label matrices.""" |
|
22
|
|
|
# Generate label vector |
|
23
|
1 |
|
y = np.hstack((np.ones((10,))*0, |
|
24
|
|
|
np.ones((10,))*1, |
|
25
|
|
|
np.ones((10,))*2)) |
|
26
|
|
|
|
|
27
|
|
|
# Map to matrix |
|
28
|
1 |
|
Y, labels = one_hot(y) |
|
29
|
|
|
|
|
30
|
|
|
# Check for only 0's and 1's |
|
31
|
1 |
|
assert len(np.setdiff1d(np.unique(Y), [0, 1])) == 0 |
|
32
|
|
|
|
|
33
|
|
|
# Check for correct labels |
|
34
|
1 |
|
assert np.all(labels == np.unique(y)) |
|
35
|
|
|
|
|
36
|
|
|
# Check correct shape of matrix |
|
37
|
1 |
|
assert Y.shape[0] == y.shape[0] |
|
38
|
1 |
|
assert Y.shape[1] == len(labels) |
|
39
|
|
|
|
|
40
|
|
|
|
|
41
|
1 |
|
def test_regularize_matrix(): |
|
42
|
|
|
"""Test whether function regularizes matrix correctly.""" |
|
43
|
|
|
# Generate test matrix |
|
44
|
1 |
|
A = rnd.randn(3) |
|
45
|
|
|
|
|
46
|
|
|
# Check for inappropriate input argument |
|
47
|
1 |
|
with pytest.raises(ValueError): |
|
48
|
|
|
regularize_matrix(A, a=-1.0) |
|
49
|
|
|
|
wmkouw /
libTLDA
