lorepy/example_uncertainty.py at main · raeslab/lorepy · GitHub

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
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
import matplotlib.pyplot as plt
import pandas as pd
from lorepy import uncertainty_plot, feature_importance
from matplotlib.colors import ListedColormap
from sklearn.datasets import load_iris
from sklearn.svm import SVC

# Load iris dataset and convert to dataframe
iris_obj = load_iris()
iris_df = pd.DataFrame(iris_obj.data, columns=iris_obj.feature_names)

iris_df["species"] = [iris_obj.target_names[s] for s in iris_obj.target]

# Default uncertainty plot
uncertainty_plot(data=iris_df, x="sepal width (cm)", y="species", iterations=100)
plt.savefig("./docs/img/uncertainty_default.png", dpi=150)
plt.show()

# Statistics for this plot
stats = feature_importance(
    data=iris_df, x="sepal width (cm)", y="species", iterations=100
)
print(stats)

stats = feature_importance(
    data=iris_df,
    x="sepal width (cm)",
    y="species",
    iterations=100,
    mode="random_subsampling",
)
print(stats)


# Using random subsampling instead of resample to assess uncertainty
uncertainty_plot(
    data=iris_df,
    x="sepal width (cm)",
    y="species",
    iterations=100,
    subsampling_fraction=0.8,
)
plt.savefig("./docs/img/uncertainty_random_subsampling.png", dpi=150)
plt.show()

# Uncertainty plot with custom colors


colormap = ListedColormap(["red", "green", "blue"])
uncertainty_plot(
    data=iris_df,
    x="sepal width (cm)",
    y="species",
    iterations=100,
    mode="resample",
    colormap=colormap,
)
plt.savefig("./docs/img/uncertainty_custom_color.png", dpi=150)
plt.show()

# Uncertainty plot with a confounder
uncertainty_plot(
    data=iris_df,
    x="sepal width (cm)",
    y="species",
    iterations=100,
    mode="resample",
    confounders=[("petal width (cm)", 1)],
)
plt.savefig("./docs/img/uncertainty_confounder.png", dpi=150)
plt.show()

# Stats with confounder
stats = feature_importance(
    data=iris_df,
    x="sepal width (cm)",
    y="species",
    iterations=100,
    mode="resample",
    confounders=[("petal width (cm)", 1)],
)

print(stats)

# Uncertainty plot with a custom classifier
svc = SVC(probability=True)

uncertainty_plot(
    data=iris_df,
    x="sepal width (cm)",
    y="species",
    iterations=100,
    mode="resample",
    clf=svc,
)
plt.savefig("./docs/img/uncertainty_custom_classifier.png", dpi=150)
plt.show()