def run_experiment(self):
# initialize velocity and population
# we need anther class that holds current population, velocity, current transformed matrix,
# a reference for population row, population iteration, current generation
# loop thru all population rows and generate fitness
for iteration in range(1, VariableSetting.Iteration + 1):
print("iteration loop",iteration)
self.current_alpha = VariableSetting.Initial_alpha + self.alpha_scaling_factor
#we can put population initialization here
for generation in range(1, VariableSetting.Generation + 1):
print("generation loop", generation)
self.current_alpha = self.current_alpha - self.alpha_scaling_factor
self.population_r2_values = np.zeros((VariableSetting.Population_Size, 3))
self.feature_selector.fitness_matrix = []
for population_idx in range(0, VariableSetting.Population_Size):
#Test area for single model row [24- 188- 274- 302- 330- 344- 362]
self.feature_selector.population_matrix[0] = np.zeros(VariableSetting.No_of_Descriptors)
self.feature_selector.population_matrix[0][24] = 1
self.feature_selector.population_matrix[0][188] = 1
self.feature_selector.population_matrix[0][274] = 1
self.feature_selector.population_matrix[0][302] = 1
self.feature_selector.population_matrix[0][330] = 1
self.feature_selector.population_matrix[0][344] = 1
self.feature_selector.population_matrix[0][362] = 1
self.feature_selector.current_population_index = population_idx
if self.feature_selector is None:
data_inputs = self.data_manager.inputs
else:
self.run_feature_selection()
data_inputs = self.data_manager.transformed_input
self.fit_and_evaluate_model(data_inputs)
#print("Row", population_idx, "Descriptor", self.feature_selector.sel_descriptors_for_curr_population, "Test r2 value ", self.population_r2_values[population_idx][2])
if(self.population_r2_values[population_idx][2] >= VariableSetting.Required_r2_Test
and self.population_r2_values[population_idx][1] >= VariableSetting.Required_r2_Valid and
self.population_r2_values[population_idx][0] >= VariableSetting.Required_r2_Train):
FileManager.write_model_in_file(self.output_filename
, self.feature_selector.sel_descriptors_for_curr_population
, self.feature_selector.fitness_matrix[population_idx]
, self.population_r2_values[population_idx][0]
, self.population_r2_values[population_idx][1]
, self.population_r2_values[population_idx][2]
)
#Test area for single model row
exit(0)
self.feature_selector.local_best_matrix = self.feature_selector.get_local_best_matrix()
if generation == 1:
self.feature_selector.initialize_local_best_fitness_for_first_generation()
self.feature_selector.global_best_row = self.feature_selector.get_global_row()
#self.print_ones_in_array(self.feature_selector.global_best_row)
self.feature_selector.find_next_velocity()
self.feature_selector.generate_population_matrix(self.current_alpha)
self.feature_selector.current_population_index = 0
#print("lowest fitness index", np.min(self.feature_selector.fitness_matrix), np.argmin(self.feature_selector.fitness_matrix))
print("Global Row fitness", self.feature_selector.global_best_row_fitness )
def run_experiment(self):
# initialize velocity and population
# we need anther class that holds current population, velocity, current transformed matrix,
# a reference for population row, population iteration, current generation
# loop thru all population rows and generate fitness
for iteration in range(1, VariableSetting.Iteration + 1):
print("iteration loop", iteration)
self.current_alpha = VariableSetting.Initial_alpha + self.alpha_scaling_factor
#we can put population initialization here
for generation in range(1, VariableSetting.Generation + 1):
print("generation loop", generation)
self.current_alpha = self.current_alpha - self.alpha_scaling_factor
self.population_r2_values = np.zeros(
(VariableSetting.Population_Size, 3))
self.feature_selector.fitness_matrix = []
for population_idx in range(0,
VariableSetting.Population_Size):
self.feature_selector.current_population_index = population_idx
if self.feature_selector is None:
data_inputs = self.data_manager.inputs
else:
self.run_feature_selection()
data_inputs = self.data_manager.transformed_input
self.fit_and_evaluate_model(data_inputs)
#print("Row", population_idx, "Descriptor", self.feature_selector.sel_descriptors_for_curr_population, "Test r2 value ", self.population_r2_values[population_idx][2])
FileManager.write_model_in_file(
self.output_filename, self.feature_selector.
sel_descriptors_for_curr_population,
self.feature_selector.fitness_matrix[population_idx],
type(self.model),
self.population_r2_values[population_idx][0],
self.population_r2_values[population_idx][1],
self.population_r2_values[population_idx][2])
self.feature_selector.local_best_matrix = self.feature_selector.get_local_best_matrix(
)
if generation == 1:
self.feature_selector.initialize_local_best_fitness_for_first_generation(
)
self.feature_selector.global_best_row = self.feature_selector.get_global_row(
)
#self.print_ones_in_array(self.feature_selector.global_best_row)
self.feature_selector.find_next_velocity()
self.feature_selector.generate_population_matrix(
self.current_alpha)
self.feature_selector.current_population_index = 0
#print("lowest fitness index", np.min(self.feature_selector.fitness_matrix), np.argmin(self.feature_selector.fitness_matrix))
print("Global Row fitness",
self.feature_selector.global_best_row_fitness)
