From 06e86b923e69f7ddc891051dc0ba796484f94937 Mon Sep 17 00:00:00 2001
From: Farid Mohammadi <farid.mohammadi@iws.uni-stuttgart.de>
Date: Mon, 20 Jun 2022 16:54:37 +0200
Subject: [PATCH] [examples][pollution] add the bootstrapping arguments.
---
examples/pollution/test_pollution.py | 5 +++++
examples/pollution/test_valid_pollution.py | 13 +++++++++----
2 files changed, 14 insertions(+), 4 deletions(-)
diff --git a/examples/pollution/test_pollution.py b/examples/pollution/test_pollution.py
index 3abd18dcb..9e9d56525 100644
--- a/examples/pollution/test_pollution.py
+++ b/examples/pollution/test_pollution.py
@@ -118,6 +118,11 @@ if __name__ == "__main__":
# 7)EBL: Emperical Bayesian Learning
MetaModelOpts.pce_reg_method = 'FastARD'
+ # Bootstraping
+ # 1) normal 2) fast
+ MetaModelOpts.bootstrap_method = 'fast'
+ MetaModelOpts.n_bootstrap_itrs = 100
+
# Specify the max degree to be compared by the adaptive algorithm:
# The degree with the lowest Leave-One-Out cross-validation (LOO)
# error (or the highest score=1-LOO)estimator is chosen as the final
diff --git a/examples/pollution/test_valid_pollution.py b/examples/pollution/test_valid_pollution.py
index 1deea2f29..0d2a32c00 100644
--- a/examples/pollution/test_valid_pollution.py
+++ b/examples/pollution/test_valid_pollution.py
@@ -97,9 +97,9 @@ if __name__ == "__main__":
MetaModelOpts = MetaModel(Inputs)
# Select if you want to preserve the spatial/temporal depencencies
- # MetaModelOpts.dim_red_method = 'PCA'
+ MetaModelOpts.dim_red_method = 'PCA'
# MetaModelOpts.var_pca_threshold = 99.999
- # MetaModelOpts.n_pca_components = 12
+ # MetaModelOpts.n_pca_components = 5
# Select your metamodel method
# 1) PCE (Polynomial Chaos Expansion) 2) aPCE (arbitrary PCE)
@@ -118,6 +118,11 @@ if __name__ == "__main__":
# 7)EBL: Emperical Bayesian Learning
MetaModelOpts.pce_reg_method = 'FastARD'
+ # Bootstraping
+ # 1) normal 2) fast
+ MetaModelOpts.bootstrap_method = 'fast'
+ MetaModelOpts.n_bootstrap_itrs = 100
+
# Specify the max degree to be compared by the adaptive algorithm:
# The degree with the lowest Leave-One-Out cross-validation (LOO)
# error (or the highest score=1-LOO)estimator is chosen as the final
@@ -170,7 +175,7 @@ if __name__ == "__main__":
# Plot the sobol indices
if MetaModelOpts.meta_model_type != 'GPE':
- sobol_cell, total_sobol = PostPCE.sobol_indices()
+ total_sobol = PostPCE.sobol_indices()
# Compute and print RMSE error
valid_samples = np.load("data/validSet.npy")
@@ -205,7 +210,7 @@ if __name__ == "__main__":
DiscOutputOpts.add_marginals()
DiscOutputOpts.Marginals[0].name = '$\\sigma^2_{\\epsilon}$'
DiscOutputOpts.Marginals[0].dist_type = 'uniform'
- DiscOutputOpts.Marginals[0].parameters = [0.0, 5.0]
+ DiscOutputOpts.Marginals[0].parameters = [0.0, 0.1]
BayesOpts.Discrepancy = Discrepancy(DiscOutputOpts)
# Start the inference
--
GitLab