def show_solution(P, T, path): best = load_network(path) Y = best.sim(P) area = plotroc(Y, T) print("") print("Stats for cut = 0.5") [num_correct_first, num_correct_second, total_performance, num_first, num_second, missed] = stat(Y, T) plt.title(path + "\nArea = " + str(area)) plt.show()
# net = build_feedforward(8, 8, 1) epochs = 10 # best = benchmark(train_evolutionary)(net, test, validation, 10, random_range = 1) # best = benchmark(traingd_block)(net, test, validation, epochs, block_size = 10, stop_error_value = 0) com = build_feedforward_committee(size=10, input_number=8, hidden_number=8, output_number=1) print "Training evolutionary..." benchmark(train_committee)(com, train_evolutionary, inputs, targets, epochs, random_range=1) Y = com.sim(inputs) area, best_cut = get_rocarea_and_best_cut(Y, targets) [num_correct_first, num_correct_second, total_performance, num_first, num_second, missed] = stat( Y, targets, cut=best_cut ) print ( "Total number of data: " + str(len(targets)) + " (" + str(num_second) + " ones and " + str(num_first) + " zeros)" ) print ("Number of misses: " + str(missed) + " (" + str(total_performance) + "% performance)") print ("Specificity: " + str(num_correct_first) + "% (Success for class 0)") print ("Sensitivity: " + str(num_correct_second) + "% (Success for class 1)") print ("Roc Area: " + str(area) + "%") save_committee(com, "/export/home/jonask/Projects/aNeuralN/ANNs/pimatrain_gen__rocarea" + str(area) + ".anncom") print "\nTraining with gradient descent..." benchmark(train_committee)(com, traingd, inputs, targets, epochs, block_size=10, stop_error_value=0) Y = com.sim(inputs)