from CTimingEstimationResult import CTimingEstimationResult
from CAlgorithmBase import CAlgorithmBase
class CAlgorithmSinglePhoton(CAlgorithmBase):
def __init__(self, photon_count):
self.__photon_count = photon_count
@property
def algorithm_name(self):
return "Single"
@property
def photon_count(self):
return self.__photon_count
def evaluate_collection_timestamps(self, coincidence_collection):
timestamps_detector1 = np.copy(coincidence_collection.detector1.timestamps[:, self.photon_count])
timestamps_detector2 = np.copy(coincidence_collection.detector2.timestamps[:, self.photon_count])
timing_estimation_results = CTimingEstimationResult(self.algorithm_name, self.photon_count, timestamps_detector1, timestamps_detector2)
return timing_estimation_results
def evaluate_single_timestamp(self, single_event):
timestamps = np.copy(single_event.timestamps[self.photon_count])
return timestamps
CAlgorithmBase.register(CAlgorithmSinglePhoton)
assert issubclass(CAlgorithmSinglePhoton, CAlgorithmBase)
#
# for label in slices:
# datasets[label] = slice_data(data, slices[label])
# return datasets
def __train_network(self):
# climate.enable_default_logging()
neural_input = self.__event_collection.detector1.timestamps[:, :self.photon_count - 1] \
- self.__event_collection.detector1.timestamps[:, self.photon_count - 1:self.photon_count]
neural_target = np.transpose(np.matrix(self.__event_collection.detector1.interaction_time.ravel()-self.__event_collection.detector1.timestamps[:, self.photon_count - 1:self.photon_count].ravel()))
self.__neural_network = theanets.Experiment(
# Neural network for regression (sigmoid hidden, linear output)
theanets.Regressor,
# Input layer, hidden layer, output layer
layers=(self.photon_count - 1, self.__hidden_layers, 1)
)
i = 0
for train, valid in self.__neural_network.itertrain([neural_input, neural_target], optimize='rmsprop'):
sys.stdout.write('\rNeural network with %d inputs, %d hidden layers - Iteration %d: Training error: %f ' %
(self.photon_count - 1, self.__hidden_layers, i, np.sqrt(train['err']) /(np.sqrt(2)/2)))
i += 1
sys.stdout.flush()
CAlgorithmBase.register(CAlgorithmNeuralNetwork)
assert issubclass(CAlgorithmNeuralNetwork, CAlgorithmBase)
from CTimingEstimationResult import CTimingEstimationResult
from CAlgorithmBase import CAlgorithmBase
class CAlgorithmMean(CAlgorithmBase):
def __init__(self, photon_count):
self.__photon_count = photon_count
@property
def algorithm_name(self):
return "Mean"
@property
def photon_count(self):
return self.__photon_count
def evaluate_collection_timestamps(self, coincidence_collection):
timestamps_detector1 = np.mean(coincidence_collection.detector1.timestamps[:, :self.photon_count], axis=1)
timestamps_detector2 = np.mean(coincidence_collection.detector2.timestamps[:, :self.photon_count], axis=1)
timing_estimation_results = CTimingEstimationResult(self.algorithm_name, self.photon_count, timestamps_detector1, timestamps_detector2)
return timing_estimation_results
def evaluate_single_timestamp(self, single_event):
timestamps = np.mean(single_event.photon_timestamps[:self.photon_count])
return timestamps
CAlgorithmBase.register(CAlgorithmMean)
assert issubclass(CAlgorithmMean, CAlgorithmBase)
@property
def photon_count(self):
return self.__photon_count
def _calculate_coefficients(self):
corrected_timestamps = self._training_coincidence_collection.detector2.timestamps[:, :self.photon_count] - self._training_coincidence_collection.detector2.interaction_time[:, None]
covariance = np.cov(corrected_timestamps[:, :self.photon_count], rowvar=0)
unity = np.ones(self.photon_count)
inverse_covariance = np.linalg.inv(covariance)
w = np.dot(unity, inverse_covariance)
n = np.dot(w, unity.T)
self._mlh_coefficients = w / n
def evaluate_collection_timestamps(self, coincidence_collection):
current_mlh_length = len(self._mlh_coefficients)
timestamps_detector1 = np.dot(coincidence_collection.detector1.timestamps[:, :current_mlh_length], self._mlh_coefficients)
timestamps_detector2 = np.dot(coincidence_collection.detector2.timestamps[:, :current_mlh_length], self._mlh_coefficients)
timing_estimation_results = CTimingEstimationResult(self.algorithm_name, self.photon_count, timestamps_detector1, timestamps_detector2)
return timing_estimation_results
def evaluate_single_timestamp(self, single_event):
return np.dot(single_event.photon_timestamps[:len(self._mlh_coefficients)], self._mlh_coefficients)
def print_coefficients(self):
self._calculate_coefficients()
print(self._mlh_coefficients)
CAlgorithmBase.register(CAlgorithmBlue)
assert issubclass(CAlgorithmBlue, CAlgorithmBase)
@property
def photon_count(self):
return self.__photon_count
def _calculate_coefficients(self):
corrected_timestamps = self._training_coincidence_collection.detector1.timestamps[:, :self.photon_count] - self._training_coincidence_collection.detector2.timestamps[:, :self.photon_count]
covariance = np.cov(corrected_timestamps[:, :self.photon_count], rowvar=0)
unity = np.ones(self.photon_count)
inverse_covariance = np.linalg.inv(covariance)
w = np.dot(unity, inverse_covariance)
n = np.dot(w, unity.T)
self._mlh_coefficients = w / n
def evaluate_collection_timestamps(self, coincidence_collection):
current_mlh_length = len(self._mlh_coefficients)
timestamps_detector1 = np.dot(coincidence_collection.detector1.timestamps[:, :current_mlh_length], self._mlh_coefficients)
timestamps_detector2 = np.dot(coincidence_collection.detector2.timestamps[:, :current_mlh_length], self._mlh_coefficients)
timing_estimation_results = CTimingEstimationResult(self.algorithm_name, self.photon_count, timestamps_detector1, timestamps_detector2)
return timing_estimation_results
def evaluate_single_timestamp(self, single_event):
return np.dot(single_event.photon_timestamps[:len(self._mlh_coefficients)], self._mlh_coefficients)
def print_coefficients(self):
self._calculate_coefficients()
print(self._mlh_coefficients)
CAlgorithmBase.register(CAlgorithmBlueDifferential)
assert issubclass(CAlgorithmBlueDifferential, CAlgorithmBase)
# Calcul des coefficients pour le detecteur 1
corrected_timestamps = self._training_coincidence_collection.detector1.timestamps[:, :self.photon_count] - timestamps_detector2[:,None]
covariance = np.cov(corrected_timestamps[:, :self.photon_count], rowvar=0)
unity = np.ones(self.photon_count)
inverse_covariance = np.linalg.inv(covariance)
w = np.dot(unity, inverse_covariance)
n = np.dot(w, unity.T)
self._mlh_coefficients = w / n
def evaluate_collection_timestamps(self, coincidence_collection):
current_mlh_length = len(self._mlh_coefficients)
timestamps_detector1 = np.dot(coincidence_collection.detector1.timestamps[:, :current_mlh_length], self._mlh_coefficients)
timestamps_detector2 = np.dot(coincidence_collection.detector2.timestamps[:, :current_mlh_length], self._mlh_coefficients)
timing_estimation_results = CTimingEstimationResult(self.algorithm_name, self.photon_count, timestamps_detector1, timestamps_detector2)
return timing_estimation_results
def evaluate_single_timestamp(self, single_event):
return np.dot(single_event.photon_timestamps[:len(self._mlh_coefficients)], self._mlh_coefficients)
def print_coefficients(self):
self._calculate_coefficients()
print(self._mlh_coefficients)
CAlgorithmBase.register(CAlgorithmBlueExpectationMaximisation)
assert issubclass(CAlgorithmBlueExpectationMaximisation, CAlgorithmBase)
