MRF.Offline package

Submodules

MRF.Offline.Data_class module

class MRF.Offline.Data_class.Data_class(training_parameters, noise_type, noise_level, minPD, maxPD, nb_files, path_files, CRBrequired=False)

Bases: MRF.BaseData_class.BaseData_class

Class allowing to deal with the importation of the precomputed fingerprints.

load_CRBs(num)

Load the file number ‘num’ containing precomputed fingerprints previously saved in the folder ‘loading_data’.

load_data(num)

Load the file number ‘num’ containing precomputed fingerprints previously saved in the folder ‘loading_data’.

load_data_from_web(num)

Load the file number ‘num’ containing precomputed fingerprints from the web.

load_urls()

Transform the text file containing the urls associated to the precomputed fingerprints.

sample()

Define the sampling strategy used to built the precomputed fingerprints files. This method is only informative and will not be used in this offline framework.

MRF.Offline.Network module

class MRF.Offline.Network.Network(name_model, loss, training_parameters, save_name, data_class, validation_settings, projection=None)

Bases: MRF.BaseNetwork.BaseNetwork, MRF.Offline.Performances.Performances

Class defining the whole neural network for training. The train method will use offline computed fingerprints.

dico_save()

Save the results and the settings of the training.

train(lr=0.001, nameoptimizer='Adam')

Launch the training using the parameter lr as learning rate.

MRF.Offline.Performances module

class MRF.Offline.Performances.Performances(validation_settings)

Bases: object

Class designed to handle the computations and the definition of the validation loss and errors.

compute_relative_errors(estimations_validation, parameters, size)

dico_save()

Save the parameters of the instance of the class Performances.

init_validation()

Define the validation dataset.

loss_function(outputs, params, size)

validation_step(estimations_validation)

Compute the loss and the relative errors on the parameters on the validation dataset. The parameter ‘estimation_validation’ represents the estimation of the network for the parameters on the validation dataset.