ML-545-Project/optimize_NN.py at master · GDhir/ML-545-Project · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
#%%
import numpy as np
from NN_airfoil import NeuralAirfoil
from sklearn.preprocessing import MinMaxScaler
from genetic_evolve import *

def read_data():
    """
    xy0, xy1, xy2 represent files corresponding to three outputs (CL, CM and CD)
    xy0:
    x: (x_geo (7 thickness, 7 cmaber) Ma, alpha),
    y: (CL), CD, CM,
    pCL/px, pCM/px, pCD/px (??? not sure about the order)
    """
    np.random.seed(0)
    xy0 = np.loadtxt('M0CFDdata.txt')
    xy1 = np.loadtxt('M1CFDdata.txt')
    xy2 = np.loadtxt('M2CFDdata.txt')
    xy = np.concatenate((np.concatenate((xy0, xy1)), xy2))
    #shuffle the data, because it has some pattern
    np.random.shuffle(xy)
    dim = 16
    x = xy[:, :dim]
    y = xy[:, dim:dim+3]
    N_train = np.int(len(y[:,0])*0.8) #Number of train data
    y_train_ok = y[:N_train,:]
    y_test_ok = y[N_train:,:]
    scaler = MinMaxScaler()
    x_train = x[:N_train,:]
    np.savetxt('x_train.txt',x_train)
    np.savetxt('y_train.txt', y_train_ok)
    x_train = scaler.fit_transform(x_train) #normalize the X_train
    x_test = scaler.transform(x[N_train:,:]) #normalize the X_test
    y_train_n = scaler.fit_transform(y[:N_train,:]) #normalized data
    y_test_n = scaler.transform(y[N_train:,:]) #normalized data
    return x_train, y_train_n, x_test, y_test_n, y_train_ok, y_test_ok

X_train, y_train, X_test, y_test, y_train_ok, y_test_ok = read_data() #Read de data

Cl_train = np.reshape(y_train[:,0],[-1,1])
Cl_test = np.reshape(y_test[:,0],[-1,1])
Cl_ok = np.reshape(y_train_ok[:,0],[-1,1])
Cl_test_ok = np.reshape(y_test_ok[:,0],[-1,1])

Cm_train = np.reshape(y_train[:,2],[-1,1])
Cm_test = np.reshape(y_test[:,2],[-1,1])
Cm_ok = np.reshape(y_train_ok[:,2],[-1,1])
Cm_test_ok = np.reshape(y_test_ok[:,2],[-1,1])

Cd_train = np.reshape(y_train[:,1],[-1,1])
Cd_test = np.reshape(y_test[:,1],[-1,1])
Cd_ok = np.reshape(y_train_ok[:,1],[-1,1])
Cd_test_ok = np.reshape(y_test_ok[:,1],[-1,1])
#%%

def obj_fun(X):
    start_time = time.time()
    learning_rate = X[0]
    n_hiddenlayers = int(X[1])
    n_neurons = np.zeros((n_hiddenlayers))
    for i in range(0,n_hiddenlayers):
        n_neurons[i] = int(X[i+2])

    epochs = 200
    verbosity = False
    #create NN
    model = NeuralAirfoil(N_hlayers=n_hiddenlayers, n_neur=n_neurons, learning_rate=learning_rate, num_epochs=epochs)
    #Train the NN
    model.train_NN(X_train, Cd_train, X_test, Cd_test, Cd_ok, Cd_test_ok, verbosity=verbosity, tolerance = 2.0E-5)
    last_mse = model.training_cost[-1]
    elapsed_time = time.time() - start_time

    totalcost = last_mse[0]/1e-6 + elapsed_time/60.0 + n_hiddenlayers + sum(n_neurons)
    outputs=[ last_mse[0], elapsed_time, totalcost]
    # outputs.append(totalcost)
    # outputs.append(last_mse)
    # outputs.append(elapsed_time)
    # print(outputs)
    return outputs

l_r = 0.001
neur = 60
layers = 2

num_cores = multiprocessing.cpu_count()

evolution = Genetic_Opt(population_size= num_cores*2 - 1, ngenes=8, datatype = [float, int,int,int,int,int,int, int], lbound = [0.00001, 1, 1, 1, 1, 1, 1, 1], ubound = [0.1, 6, 200, 200, 200, 200, 200, 200], fitfun=obj_fun, max_mutation_rate=1, max_mutation_size=5, history_filename='cd_train_history_multilayer10.csv')
evolution.initial_population([0.001, 3, 30,30,30,30,30,30])
evolution.evolve(50)


# last = obj_fun(l_r,neur,layers)
# print(last)