From 35d0b3db91fdd6613df2130a2cf960b02483111a Mon Sep 17 00:00:00 2001
From: Danilo Ferreira de Lima <danilo.enoque.ferreira.de.lima@xfel.de>
Date: Tue, 31 Jan 2023 17:49:28 +0100
Subject: [PATCH] Corrected PCA fit.
---
pes_to_spec/model.py | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)
diff --git a/pes_to_spec/model.py b/pes_to_spec/model.py
index d45e477..55bd17d 100644
--- a/pes_to_spec/model.py
+++ b/pes_to_spec/model.py
@@ -600,18 +600,23 @@ class Model(TransformerMixin, BaseEstimator):
Returns: Smoothened high resolution spectrum.
"""
+ print("Fitting PCA on low-resolution data.")
x_t = self.x_model.fit_transform(low_res_data)
+ print("Fitting PCA on high-resolution data.")
y_t = self.y_model.fit_transform(high_res_data, smoothen__energy=high_res_photon_energy)
#self.fit_model.set_params(fex__gamma=1.0/float(x_t.shape[0]))
+ print("Fitting model.")
self.fit_model.fit(x_t, y_t)
# calculate the effect of the PCA
+ print("Calculate PCA unc. on high-resolution data.")
high_res = self.y_model['smoothen'].transform(high_res_data)
high_pca = self.y_model['pca'].transform(high_res)
high_pca_rec = self.y_model['pca'].inverse_transform(high_pca)
high_pca_unc = np.sqrt(np.mean((high_res - high_pca_rec)**2, axis=0, keepdims=True))
self.y_model['unc'].set_uncertainty(high_pca_unc)
+ print("Calculate PCA unc. on low-resolution data.")
low_res = self.x_model['select'].transform(low_res_data)
pca_model = self.x_model['pca']
if 'fex' in self.x_model.named_steps:
@@ -622,13 +627,16 @@ class Model(TransformerMixin, BaseEstimator):
self.x_model['unc'].set_uncertainty(low_pca_unc)
# for consistency check per channel
+ print("Calculate PCA per channel on low-resolution data.")
selection_model = self.x_model['select']
low_res = selection_model.transform(low_res_data, keep_dictionary_structure=True)
for channel in self.get_channels():
+ print(f"Calculate PCA on {channel}")
low_pca = self.channel_pca_model[channel].named_steps["pca"].fit_transform(low_res[channel])
low_pca_rec = self.channel_pca_model[channel].named_steps["pca"].inverse_transform(low_pca)
- low_pca_unc = np.mean(np.sqrt(np.mean((low_res - low_pca_rec)**2, axis=1, keepdims=True)), axis=0, keepdims=True)
+ low_pca_unc = np.mean(np.sqrt(np.mean((low_res[channel] - low_pca_rec)**2, axis=1, keepdims=True)), axis=0, keepdims=True)
self.channel_pca_model[channel]['unc'].set_uncertainty(low_pca_unc)
+ print("End of fit.")
return high_res
--
GitLab