Skip to content

Commit 21a79c9

Browse files
JoschaBoehnleinJoschaBoehnlein
authoredApr 12, 2024
Add files via upload
1 parent 5c88597 commit 21a79c9

File tree

2 files changed

+144
-0
lines changed

2 files changed

+144
-0
lines changed
 

‎default_pipeline.py

+72
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,72 @@
1+
# -------------------- GENERATED WITH PHOTON WIZARD (beta) ------------------------------
2+
# PHOTON Project Folder: C:/Users/j_boeh06/Sciebo/Forschung/PythonProjects/multimodal_ER
3+
4+
5+
import pandas as pd
6+
import numpy as np
7+
from photonai.base import Hyperpipe, PipelineElement, OutputSettings, Switch, Preprocessing
8+
from photonai.optimization import Categorical, IntegerRange, FloatRange
9+
from sklearn.model_selection import RepeatedKFold, KFold
10+
11+
# Specify how results are going to be saved
12+
# Define hyperpipe
13+
hyperpipe = Hyperpipe('defaultpipelinemultimodaler',
14+
project_folder="C:/Users/j_boeh06/Sciebo/Forschung/PythonProjects/multimodal_ER",
15+
optimizer="grid_search",
16+
optimizer_params={},
17+
metrics=['mean_squared_error', 'mean_absolute_error', 'explained_variance', 'pearson_correlation',
18+
'r2'],
19+
best_config_metric="mean_squared_error",
20+
outer_cv=RepeatedKFold(n_splits=10, random_state=42, n_repeats=10),
21+
inner_cv=KFold(n_splits=10, shuffle=True, random_state=42),
22+
verbosity=1)
23+
24+
# Add Simple Imputer to impute missing
25+
hyperpipe += PipelineElement("SimpleImputer")
26+
27+
# Add preprocessing: Robust Scaler
28+
hyperpipe += PipelineElement("RobustScaler", hyperparameters={},
29+
test_disabled=False, with_centering=True, with_scaling=True)
30+
31+
# Add feature engineering via Switch
32+
# Tests three options PCA, off (when PCA disabled = true) and FSelect
33+
transformer_switch = Switch('TransformerSwitch')
34+
transformer_switch += PipelineElement("PCA", hyperparameters={'n_components':None},
35+
test_disabled=True)
36+
transformer_switch += PipelineElement("FRegressionSelectPercentile", hyperparameters={'percentile': [5, 10, 50]},
37+
test_disabled=False)
38+
hyperpipe += transformer_switch
39+
40+
# Add estimator
41+
# Defaults as defined in the model zoo default parameters: https://github.com/wwu-mmll/photonai/blob/main/photonai/base/model_zoo.py
42+
estimator_switch = Switch('EstimatorSwitch')
43+
# I could save some time here by removing 1e-8 (0.00000001) and 1e8 (100,000,000) without losing too much
44+
estimator_switch += PipelineElement("SVR", hyperparameters={'C': [1e-8, 1e-6, 1e-4, 1e-2, 1, 1e2, 1e4, 1e6, 1e8], 'kernel': ['linear', 'rbf']},
45+
max_iter=1000)
46+
estimator_switch += PipelineElement("RandomForestRegressor", hyperparameters={'max_features': ['sqrt', 'log2'],
47+
'min_samples_leaf': [0.01, 0.1, 0.2]})
48+
hyperpipe += estimator_switch
49+
50+
# Load data
51+
# Adjust this to the modality and its data sheet!
52+
df = pd.read_excel('C:/Users/j_boeh06/Sciebo/Forschung/PythonProjects/multimodal_ER/features.xlsx')
53+
X = np.asarray(df.iloc[:, 1:100])
54+
y = np.asarray(df.iloc[:, 0])
55+
56+
# Fit hyperpipe
57+
hyperpipe.fit(X, y)
58+
59+
# Feature Importances
60+
r = hyperpipe.get_permutation_feature_importances(n_repeats=50, random_state=0)
61+
print(r)
62+
63+
for i in r["mean"].argsort()[::-1]:
64+
if r["mean"][i] - 2 * r["std"][i] > 0:
65+
# Hier Anpassungen von mir: Bezug auf df, Schreiben der entsprechenden Spalten-Header vor die Durchschnitts- und SD-Werte
66+
print(f"{df.columns[i]:<8}: "
67+
f"{r['mean'][i]:.3f}"
68+
f" +/- {r['std'][i]:.3f}")
69+
70+
# Schreiben der Feature Importances in eine CSV-Datei
71+
feature_importance = pd.DataFrame(r)
72+
r.to_csv('feature_importances.csv')

‎sanity_check_2024_03_08.py

+72
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,72 @@
1+
# -------------------- GENERATED WITH PHOTON WIZARD (beta) ------------------------------
2+
# PHOTON Project Folder: C:/Users/j_boeh06/Sciebo/Forschung/PythonProjects/multimodal_ER
3+
4+
5+
import pandas as pd
6+
import numpy as np
7+
from photonai.base import Hyperpipe, PipelineElement, OutputSettings, Switch, Preprocessing
8+
from photonai.optimization import Categorical, IntegerRange, FloatRange
9+
from sklearn.model_selection import KFold
10+
11+
# Specify how results are going to be saved
12+
# Define hyperpipe
13+
hyperpipe = Hyperpipe('sanity_check_kevin',
14+
project_folder="C:/Users/j_boeh06/Sciebo/Forschung/PythonProjects/multimodal_ER/sanity_check",
15+
optimizer="grid_search",
16+
optimizer_params={},
17+
metrics=['mean_squared_error', 'mean_absolute_error', 'explained_variance', 'pearson_correlation',
18+
'r2'],
19+
best_config_metric="mean_squared_error",
20+
outer_cv=KFold(n_splits=10, shuffle=True, random_state=42),
21+
inner_cv=KFold(n_splits=10, shuffle=True, random_state=42),
22+
verbosity=1)
23+
24+
25+
26+
27+
# Add preprocessing: Robust Scaler
28+
hyperpipe += PipelineElement("RobustScaler", hyperparameters={},
29+
test_disabled=False, with_centering=True, with_scaling=True)
30+
31+
# Add feature engineering via Switch
32+
# Tests three options PCA, off (when PCA disabled = true) and FSelect
33+
transformer_switch = Switch('TransformerSwitch')
34+
transformer_switch += PipelineElement("PCA", hyperparameters={'n_components':None},
35+
test_disabled=True)
36+
transformer_switch += PipelineElement("FRegressionSelectPercentile", hyperparameters={'percentile': [5, 10, 50]},
37+
test_disabled=False)
38+
hyperpipe += transformer_switch
39+
40+
# Add estimator
41+
# Defaults as defined in the model zoo default parameters: https://github.com/wwu-mmll/photonai/blob/main/photonai/base/model_zoo.py
42+
estimator_switch = Switch('EstimatorSwitch')
43+
# I could save some time here by removing 1e-8 (0.00000001) and 1e8 (100,000,000) without losing too much
44+
estimator_switch += PipelineElement("SVR", hyperparameters={'C': [1e-8, 1e-6, 1e-4, 1e-2, 1, 1e2, 1e4, 1e6, 1e8], 'kernel': ['linear', 'rbf']},
45+
max_iter=1000)
46+
estimator_switch += PipelineElement("RandomForestRegressor", hyperparameters={'max_features': ['sqrt', 'log2'],
47+
'min_samples_leaf': [0.01, 0.1, 0.2]})
48+
hyperpipe += estimator_switch
49+
50+
# Load data
51+
# Adjust this to the modality and its data sheet!
52+
df = pd.read_excel('C:/Users/j_boeh06/Sciebo/Forschung/PythonProjects/multimodal_ER/sanity_check/d_feat_sanity_with_labels.xlsx')
53+
X = np.asarray(df.iloc[:, 1:100])
54+
y = np.asarray(df.iloc[:, 0])
55+
56+
# Fit hyperpipe
57+
hyperpipe.fit(X, y)
58+
59+
# Feature Importances
60+
r = hyperpipe.get_permutation_feature_importances(n_repeats=50, random_state=0)
61+
print(r)
62+
63+
for i in r["mean"].argsort()[::-1]:
64+
if r["mean"][i] - 2 * r["std"][i] > 0:
65+
#Hier Anpassungen von mir: Bezug auf df, Schreiben der entsprchenden Spalten-Header vor die Durchschnitts- und SD-Werte
66+
print(f"{df.columns[i]:<8}: "
67+
f"{r['mean'][i]:.3f}"
68+
f" +/- {r['std'][i]:.3f}")
69+
70+
# Schreiben der Feature Importances in eine CSV-Datei
71+
feature_importance = pd.DataFrame(r)
72+
feature_importance.to_csv('feature_importances.csv')

0 commit comments

Comments
 (0)
Please sign in to comment.