From 8eb2d8be399cdb716f0a35d3e001910b280cb710 Mon Sep 17 00:00:00 2001
From: Danilo Ferreira de Lima <danilo.enoque.ferreira.de.lima@xfel.de>
Date: Mon, 19 Dec 2022 18:39:53 +0100
Subject: [PATCH] Save properties.
---
pes_to_spec/model.py | 37 +++++++++++++++++++++++--------------
1 file changed, 23 insertions(+), 14 deletions(-)
diff --git a/pes_to_spec/model.py b/pes_to_spec/model.py
index af421a4..fa88929 100644
--- a/pes_to_spec/model.py
+++ b/pes_to_spec/model.py
@@ -9,7 +9,7 @@ from scipy.optimize import fmin_l_bfgs_b
from sklearn.decomposition import PCA, IncrementalPCA
from sklearn.model_selection import train_test_split
-import logging
+from time import time_ns
import matplotlib.pyplot as plt
@@ -84,6 +84,18 @@ class Model(object):
# smoothing of the SPEC data in eV
self.high_res_sigma = high_res_sigma
+ def parameters(self) -> Dict[str, Any]:
+ """
+ Dump parameters as a dictionary.
+ """
+ return dict(channels=self.channels,
+ n_pca_lr=self.n_pca_lr,
+ n_pca_hr=self.n_pca_hr,
+ high_res_sigma=self.high_res_sigma,
+ tof_start=self.tof_start,
+ delta_tof=self.delta_tof,
+ validation_size=self.validation_size)
+
def preprocess_low_res(self, low_res_data: Dict[str, np.ndarray]) -> np.ndarray:
"""
Get a dictionary with the channel names for the inut low resolution data and output
@@ -181,34 +193,23 @@ class Model(object):
self.high_res_photon_energy = high_res_photon_energy
- print("Find peaks.")
# if the prompt peak has not been given, guess it
if self.tof_start is None:
self.tof_start = self.estimate_prompt_peak(low_res_data)
- print("Prompt at", self.tof_start)
- print("Pre-processing low")
low_res = self.preprocess_low_res(low_res_data)
- print("Pre-processing high")
high_res = self.preprocess_high_res(high_res_data, high_res_photon_energy)
# fit PCA
- print("PCA low", low_res.shape)
low_pca = self.lr_pca.fit_transform(low_res)
- print("PCA high")
high_pca = self.hr_pca.fit_transform(high_res)
- print("Split")
# split in train and test for PCA uncertainty evaluation
low_pca_train, low_pca_test, high_pca_train, high_pca_test = train_test_split(low_pca, high_pca, test_size=self.validation_size, random_state=42)
# fit the linear model
- print("Fit")
self.fit_model.fit(low_pca_train, high_pca_train, low_pca_test, high_pca_test)
- print("PCA unc")
high_pca_rec = self.hr_pca.inverse_transform(high_pca)
self.high_pca_unc = np.sqrt(np.mean((high_res - high_pca_rec)**2, axis=0, keepdims=True))
- print("Done")
-
return high_res
def predict(self, low_res_data: Dict[str, np.ndarray]) -> np.ndarray:
@@ -228,8 +229,13 @@ class Model(object):
high_pca = self.fit_model.predict(low_pca)
high_res_predicted = self.hr_pca.inverse_transform(high_pca["Y"])
n_high_res_features = high_res_predicted.shape[1]
- high_res_unc = self.hr_pca.inverse_transform(high_pca["Y"] + high_pca["Y_eps"]) - high_res_predicted
- result = np.stack((high_res_predicted, high_res_unc, np.broadcast_to(self.high_pca_unc, (n_trains, n_high_res_features))), axis=2)
+ high_res_unc = (self.hr_pca.inverse_transform(high_pca["Y"] + high_pca["Y_eps"])
+ - high_res_predicted)
+ result = np.stack((high_res_predicted,
+ high_res_unc,
+ np.broadcast_to(self.high_pca_unc,
+ (n_trains, n_high_res_features))),
+ axis=2)
return result
def save(self, filename: str, lr_pca_filename: str, hr_pca_filename: str):
@@ -243,6 +249,7 @@ class Model(object):
"""
with h5py.File(filename, 'w') as hf:
d = self.fit_model.as_dict()
+ d.update(self.parameters())
for key, value in d.items():
if isinstance(value, int):
hf.attrs[key] = value
@@ -266,6 +273,8 @@ class Model(object):
d = {k: hf[k][()] for k in hf.keys()}
d.update({k: hf.attrs[k] for k in hf.attrs})
self.fit_model.from_dict(d)
+ for key in self.parameters().keys():
+ setattr(self, key, d[key])
self.lr_pca = joblib.load(lr_pca_filename)
self.hr_pca = joblib.load(hr_pca_filename)
--
GitLab