def calculate(inlayer3, weights1, bias1, weights2, bias2, error1, goku, q): hlayer1 = [0, 0, 0, 0, 0, 0, 0, 0] hlayer1 = nn.cal_hid(inlayer3, weights1, hlayer1, bias1) y = nn.cal_out(hlayer1, weights2, bias2) error = nn.cal_error(y, t) if error1 > error: goku = q error1 = error return error1, goku
def nn1(weights1, weights2, bias1, bias2): import neural_network as nn f = open('move1.txt', 'r') inlayer = [0, 0, 0, 0, 0, 0, 0, 0, 0] # bias1 = [] # Bias between Input and Hidden Layer # bias2 = 0.0 # Bias between Hidden layer and Output olayerlayer = 0.0 # weights1 = [] # input to hidden1 # weights2 = [] # hidden1 to hidden2 y = 0.0 error = 0.0 cost = 0.0 ''' weights1 = nn.w1(weights1) weights2 = nn.w2(weights2) bias1, bias2 = nn.bias(bias1, bias2) weights1, weights2, bias1, bias2 = test.w()''' for k in range(50): cost = 0.0 f.seek(0, 0) for j in range(100000): count1 = f.readline() count = str() for i in range(len(count1) - 1): count += count1[i] count = count.split(',') l = [] for i in range(len(count)): l.append(int(count[i])) # print(l) # print(l) inlayer = [] for i in range(9): inlayer.append(l[i]) t = l[9] # print(inlayer, t) hlayer1 = [0, 0, 0, 0, 0, 0, 0, 0] hlayer1 = nn.cal_hid(inlayer, weights1, hlayer1, bias1) y = nn.cal_out(hlayer1, weights2, bias2) error = nn.cal_error(y, t) if k == 49: errors1.append(error) cost += error weights1, weights2, bias1, bias2 = nn.Adjust_weights( inlayer, weights1, bias1, hlayer1, weights2, bias2, y, t) del (count, count1) print("Epoch ", k + 1, " : ") print("\tCost : ", cost) errors.append(cost) f.close() return weights1, weights2, bias1, bias2
count = str() for i in range(len(count1) - 1): count += count1[i] count = count.split(',') l = [] for i in range(len(count)): l.append(int(count[i])) # print(l) inlayer = [] for i in range(9): inlayer.append(l[i]) t = l[9] # print(inlayer, t) hlayer1 = [0, 0, 0, 0, 0, 0, 0, 0] hlayer1 = nn.cal_hid(inlayer, weights1, hlayer1, bias1) y, olayer = nn.cal_out(hlayer1, weights2, y, bias2) error = nn.cal_error(y, t) if k == 49999: errors1.append(error) cost += error weights1, weights2, bias1, bias2 = nn.Adjust_weights( inlayer, weights1, bias1, hlayer1, weights2, bias2, y, t) del (count, count1) print("Epoch ", k, " : ") print("Cost : ", cost) errors.append(cost) if k % 10 == 0: print(weights1, '\n', weights2, '\n', bias1, '\n', bias2) plt.figure(1)