+ "_E"
+ str(epochs)
+ "_10fold"
+ str(each_fold)
)
file_prefix += "_L" + str(1) + "_H" + str(hidden_size)
print "--------------"
print "Prefijo de experimento:", file_prefix
test_set = appraisals_random_sorted[each_fold * data_slice : (each_fold + 1) * data_slice]
training_set = (
appraisals_random_sorted[: each_fold * data_slice]
+ appraisals_random_sorted[(each_fold + 1) * data_slice :]
)
# entrenamiento de una red neuronal de una capa
training_stats, trained_network = ann_engine.train_ann(
training_set, data_manager.input_fields, hidden_size, epochs
)
pickle.dump(trained_network, open(file_prefix + "ANN_pybrain", "wb"))
exp_stats = {
"Clave_experimento": file_prefix,
"Archivo_CSV_ejemplos": file_name_examples,
"CP": filter_string_msg,
}
print "EXPERIMENT STATS:", exp_stats
print "TRAINING STATS:", training_stats
# activacion de la red con los avaluos de prueba
activations_stats = ann_engine.activate_network(trained_network, training_set, data_manager.input_fields)
print "----------------------" # selecciona el 80-20 para entrenamiento y pruebas percent_80 = int(len(appraisals)*0.8) training_app = appraisals[:percent_80] test_app = appraisals[percent_80:] # entrenamiento de una red neuronal de una capa trained_network = ann_engine.train_ann(training_app, None, hidden_size = 50, epochs = 1000) #trained_network = ann_engine.train_ann_multihidden(training_app, None, hidden_size = 50, epochs = 1500) # activacion de la red con los avaluos de prueba ann_engine.activate_network(trained_network, test_app)
