def differentiate(point):
# This will be the output of the function
partials = []
# The -1 index of 'point' is the label
label = point[-1]
# Prediction from neural network
prediction = neural_network(point, network_weights, bias)
# Create the 'weight symbols', which will be used
# to represent the variables in the cost equation
weight_symbols = []
# Represents the bias term
b = sym.symbols('b')
for i in range(len(network_weights)):
symbol = 'w' + str(i)
symbol = sym.symbols(symbol)
weight_symbols.append(symbol)
# Adds all the network weights and bias to the prediction,
# which then goes into the sigmoid function
pred = 0
for i in range(len(network_weights)):
pred += weight_symbols[i] * network_weights[i]
pred += b
# Take sigmoid of the prediction
sig = 1 / (1 + sym.exp(-pred))
#print(sig)
# Final cost function
f_cost = (sig - label)**2
#print(len(network_weights))
# Take partial derivative of final cost function
# with respect to all the weights
for i in range(len(network_weights) + 1):
# The -1 index of the output will be the
# partial with respect to the bias term
if (i == len(network_weights)):
alg_dcost_dweight = sym.diff(f_cost, b)
# Algebraic expression for derivative
else:
alg_dcost_dweight = sym.diff(f_cost, weight_symbols[i])
alg_dcost_dweight = alg_dcost_dweight.subs(b, bias)
# Simplified number for derivative
simp_dcost_dweight = alg_dcost_dweight.subs({
weight_symbols[j]: network_weights[j]
for j in range(len(network_weights))
})
answer = simp_dcost_dweight.evalf()
partials.append(answer)
# Take partial derivative of final cost function
# with respect to the bias term
#print(partials)
#for i in range(len(network_weights)-1):
return partials
############### Name: Shubham Pareek ############ ############### UBID: spareek ############ from logistic_regression import * from linear_regression import * from neural_network import * from preprocessing import * X1, y1 = get_feature_matrix(data='hod', method='concatenate') X2, y2 = get_feature_matrix(data='hod', method='subtract') X3, y3 = get_feature_matrix(data='gsc', method='concatenate') X4, y4 = get_feature_matrix(data='gsc', method='subtract') logistic_regression(X1, y1) logistic_regression(X2, y2) logistic_regression(X3, y3) logistic_regression(X4, y4) linear_regression(X1, y1) linear_regression(X2, y2) linear_regression(X3, y3) linear_regression(X4, y4) neural_network(X1, y1) neural_network(X2, y2) neural_network(X3, y3) neural_network(X4, y4)
window.pushButton_3.clicked.connect(window.run_pso)
window.pushButton_5.clicked.connect(window.auto_fill)
window.pushButton_7.clicked.connect(window.shut_down_plot)
window.pushButton_8.clicked.connect(window.go_on_nn)
window.pushButton_10.clicked.connect(window.retrieve)
window.pushButton_11.clicked.connect(window.suspend_pso)
window.pushButton_12.clicked.connect(window.go_on_pso)
window.pushButton_9.clicked.connect(window.reset_nn)
window.pushButton_13.clicked.connect(window.suspend_nn)
window.pushButton_13.clicked.connect(window.pause_plot)
t = time_e() # 相当于初始化__init__函数
t.update_time.connect(window.handle_display_time)
t.start() # 相当于执行run()函数
nn = neural_network()
nn.update_msg.connect(window.handle_display)
nn.update_info.connect(window.handle_display)
nn.update_acc.connect(window.handle_display_acc)
nn.update_top_acc.connect(window.handle_display_top_acc)
nn.stop_time.connect(t.handle_stop_signal)
nn.update_plot_acc.connect(window.handle_display_plot)
nn.start() # 单独在这里运行线程，不能放在主进程里，否则会被视为主进程的一部分（？）
cbr_system = CBR()
cbr_system.transmit_data.connect(window.show_table)
cbr_system.best_case.connect(window.set_init_global_best)
cbr_system.transmit_span.connect(window.set_span)
cbr_system.best_distance.connect(window.set_best_distance)
cbr_system.start()
#ACCURACY :
#Rule : must be a float, the minimum accuracy at which the algorithm will stop iteratig (represent a minimum fitness,
#that has to be reached by the best individual).
ACCURACY = 0.01 #default : 0.01
if len(sys.argv) != 3:
print(
'Please notice that this program need 2 arguments, the first being the path to the csv file containing the '
'samples. The second being the number of output of these sample.')
exit()
samples = read_data_obj(filename=sys.argv[1],
number_of_output=int(sys.argv[2]))
reseau = neural_network(training_data=samples,
function_by_layer=FUNCTION_TO_USE_BY_LAYER,
layer_structure=LAYER_STRUCTURE,
max_generations=MAX_GENERATIONS,
number_of_individuals=NUMBER_OF_INDIVIDUALS,
number_of_children=NUMBER_OF_CHILDREN)
print(reseau)
#RESULTS OUTPUT :
results = reseau.train()
print('Success :', results[1])
print('Nb of generation :', results[2])
print('Best individual :\n', results[0])
weight_output = []
for a in reseau.produce_output():
weight_output.append(a)
print('Output for this weight :', weight_output)
print('Ideal output :', reseau.goal_output)
from initialize_data import *
from initialize_weights import *
from training_loop import *
from sigmoid import *
from neural_network import *
from partial_derivatives import *
# Import dependencies
import sympy as sym
import time
start_time = time.time()
# Train neural network
train(45)
# Test neural network
accuracy = 0
for j in range(45, 90):
# Creates an array with all the features of the given
# data point. The -1 index of 'point' is the label
point = []
for k in range(number_of_features):
point.append(testing_labels[k][j])
pred = neural_network(point, network_weights, bias)
print(point)
print("Result for " + str(j) + " is: " + str(pred) + "\n")
if round(pred) == point[-1]:
accuracy += 1
print("The accuracy of the neural network is: " + str(accuracy / 45 * 100) +
"%")
print("--- %s seconds ---" % (time.time() - start_time))
from evolutive_algorithm_class import *
from neural_network import *
from functions import *
from sample_production import *
import graphical
samples = read_data_obj(
filename=
r'C:\Users\Crowbar\PycharmProjects\BIA\trainNNwithEvolutiveAlgo\samples.txt',
number_of_output=1)
reseau = neural_network(training_data=samples,
function_by_layer=[function_sigmoid, function_sigmoid],
layer_structure=[2, 1],
max_generations=2000,
range_w=[-10, 10])
print(reseau)
results = reseau.train()
print('Success :', results[1])
print('Nb of generation :', results[2])
print('Best individual :\n', results[0])
weight_output = []
for a in reseau.produce_output():
weight_output.append(a)
print('Output for this weight :', weight_output)
print('Ideal output :', reseau.goal_output)
graphical.display_evolution(results[3],
'Evolution of the global MSE for each generation')
graphical.display_evolution(results[4],
'Evolution of the best FF for each generation')