DigitRecognitionPython/DigitRecognitionPython.py at master · sn2234/DigitRecognitionPython · 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
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214

import sys
import numpy as np
import matplotlib.image as img
import matplotlib.pyplot as plt
from sklearn.metrics import accuracy_score


import DataModel
import Train
import NN_1HL
import SimpleNN2

def findBestRegularization(s, x_sub, y_sub):
    regs = np.linspace(0, 10, 20)
    reg_acc_cv = []
    reg_acc_train = []
    max_acc = 0
    best_reg = 0

    for r in regs:
        th1, th2 = Train.trainSciPy2(s, x_sub, y_sub, r)

        acc_cv = accuracy_score(y_cv, [SimpleNN2.predictClass(s, th1, th2, w) for w in x_cv])
        acc_train = accuracy_score(y_sub, [SimpleNN2.predictClass(s, th1, th2, w) for w in x_sub])
        reg_acc_cv.append(acc_cv)
        reg_acc_train.append(acc_train)

        if max_acc < acc_cv:
            max_acc = acc_cv
            best_reg = r


        print("Validating regularization parameter [{0}]; Train accuracy: [{1}] CV accuracy: [{2}]"
              .format(r, acc_train, acc_cv))

    print("Best reg param: {0} with accuracy on CV dataset: {1}".format(best_reg, max_acc))

    plt.plot(regs, reg_acc_cv);plt.plot(regs, reg_acc_train)
    plt.show()

    return best_reg

def test1():
    (x, y) = DataModel.loadData("..\\train.csv")

    (x_train, x_cv, y_train, y_cv) = DataModel.splitData(x, y)

    x_sub = x_train[:500,:]
    y_sub = y_train[:500]

    s = SimpleNN.SimpleNN([784, 70, 10])

    #s = Train.trainGradientDescent(s, x_sub, y_sub, 5)
    s = Train.trainSciPy(s, x_sub, y_sub, 5)
    acc_cv = accuracy_score(y_cv, [s.predictClass(w) for w in x_cv])
    print("Accuracy on CV set: {0}", acc_cv)

def test2():
    (x, y) = DataModel.loadData("..\\train.csv")

    y = y.astype(int)

    (x_train, x_cv, y_train, y_cv) = DataModel.splitData(x, y)

    x_sub = x_train[:500,:]
    y_sub = y_train[:500]

    s = NN_1HL.NN_1HL(reg_lambda = 1, opti_method = 'CG')
    s.fit(x_sub, y_sub)

    acc_cv = accuracy_score(y_cv, [s.predict(w) for w in x_cv])
    print("Accuracy on CV set: {0}", acc_cv)

def test3():
    (x, y) = DataModel.loadData("..\\train.csv")

    (x_train, x_cv, y_train, y_cv) = DataModel.splitData(x, y)

    x_sub = x_train[:20000,:]
    y_sub = y_train[:20000]

    s = SimpleNN2.NeuralNetConfig(784, 70, 10)

    regLambda = 6.84
    #s = Train.trainGradientDescent(s, x_sub, y_sub, 5)
    th1, th2 = Train.trainSciPy2(s, x_sub, y_sub, regLambda)
    #th1, th2 = Train.trainGradientDescent2(s, x_sub, y_sub, 5)

    acc_cv = accuracy_score(y_cv, [SimpleNN2.predictClass(s, th1, th2, w) for w in x_cv])
    print("Accuracy on CV set: {0}".format(acc_cv))

def compareImplementations():
    (x, y) = DataModel.loadData("..\\train.csv")

    y = y.astype(int)

    (x_train, x_cv, y_train, y_cv) = DataModel.splitData(x, y)

    x_sub = x_train[:500,:]
    y_sub = y_train[:500]

    s_my = SimpleNN.SimpleNN([784, 70, 10])
    s_t = NN_1HL.NN_1HL(reg_lambda = 1, opti_method = 'CG')

    np.random.seed(123)

    thetas = [s_t.rand_init(784,70), s_t.rand_init(70, 10)]

    cost_t = s_t.function(s_t.pack_thetas(thetas[0].copy(), thetas[1].copy()), 784, 70, 10, x_sub, y_sub, 10)
    grad_t = s_t.function_prime(s_t.pack_thetas(thetas[0], thetas[1]), 784, 70, 10, x_sub, y_sub, 10)
    print(cost_t, np.sum(grad_t));

    cost_my = s_my.computeCost(s_my.combineTheta(thetas.copy()), x_sub, y_sub, 10)
    grad_my = s_my.computeGrad(s_my.combineTheta(thetas), x_sub, y_sub, 10)

    print(cost_my, np.sum(grad_my))

def compareImplementations2():
    (x, y) = DataModel.loadData("..\\train.csv")

    y = y.astype(int)

    (x_train, x_cv, y_train, y_cv) = DataModel.splitData(x, y)

    x_sub = x_train[:500,:]
    y_sub = y_train[:500]

    s_my = SimpleNN2.NeuralNetConfig(784, 70, 10)
    s_t = NN_1HL.NN_1HL(reg_lambda = 10, opti_method = 'CG')

    np.random.seed(123)

    thetas = [s_t.rand_init(784,70), s_t.rand_init(70, 10)]

    # Check costs
    cost_t = s_t.function(s_t.pack_thetas(thetas[0].copy(), thetas[1].copy()), 784, 70, 10, x_sub, y_sub, 10)
    print("Cost test: ", cost_t)

    cost_my = SimpleNN2.computeCost(s_my, thetas[0], thetas[1], x_sub, y_sub, 10)
    print("Cost my: ", cost_my)

    # Check gradients
    grad_t = s_t.function_prime(s_t.pack_thetas(thetas[0].copy(), thetas[1].copy()), 784, 70, 10, x_sub, y_sub, 10)
    print("Grad sum test: ", np.sum(grad_t))

    grad_my1, grad_my2 = SimpleNN2.computeGrad(s_my, thetas[0], thetas[1], x_sub, y_sub, 10)
    print("Grad sum my: ", np.sum(grad_my1) + np.sum(grad_my2))

def trainFullAndSave():
    (x, y) = DataModel.loadData("..\\train.csv")

    (x_train, x_cv, y_train, y_cv) = DataModel.splitData(x, y)

    s = SimpleNN2.NeuralNetConfig(784, 70, 10)

    regLambda = 6.84

    print("Training neural network on full dataset")
    #s = Train.trainGradientDescent(s, x_sub, y_sub, 5)
    th1, th2 = Train.trainSciPy2(s, x_train, y_train, regLambda)
    #th1, th2 = Train.trainGradientDescent2(s, x_sub, y_sub, 5)

    print("Training complete, checking accuracy on CV data")

    acc_cv = accuracy_score(y_cv, [SimpleNN2.predictClass(s, th1, th2, w) for w in x_cv])
    print("Accuracy on CV set: {0}".format(acc_cv))

    SimpleNN2.saveNetwork(s, th1, th2, "..\\NeuralNetwork.bin")

def makeTestPerdictions():
    x, _ = DataModel.loadData("..\\test.csv")
    s, th1, th2 = SimpleNN2.loadNetwork("..\\NeuralNetwork.bin")

    y = [SimpleNN2.predictClass(s, th1, th2, w) for w in x]

    with open("results.csv", "w") as f:
        imageId = 1
        f.write("ImageId,Label\n")
        for i in y:
            f.write("{0},{1}\n".format(imageId, i))
            imageId = imageId + 1

#test3()


#(x, y) = DataModel.loadData("..\\train.csv")

#(x_train, x_cv, y_train, y_cv) = DataModel.splitData(x, y)

#subsetLen = 200
#x_sub = x_train[:subsetLen,:]
#y_sub = y_train[:subsetLen]

#s = SimpleNN2.NeuralNetConfig(784, 70, 10)
#bestReg = findBestRegularization(s, x_sub, y_sub)

makeTestPerdictions()

#(x, y) = DataModel.loadData("..\\train.csv")

#(x_train, x_cv, y_train, y_cv) = DataModel.splitData(x, y)

#s = SimpleNN2.NeuralNetConfig(784, 70, 10)

#regLambda = 6.84
##s = Train.trainGradientDescent(s, x_sub, y_sub, 5)
#th1, th2 = Train.trainSGD(s, x_train, y_train, regLambda)
##th1, th2 = Train.trainGradientDescent2(s, x_sub, y_sub, 5)
#costFinal = SimpleNN2.computeCost(s, th1, th2, x_train, y_train, regLambda)
#print("Final cost: {0}".format(costFinal))

#acc_cv = accuracy_score(y_cv, [SimpleNN2.predictClass(s, th1, th2, w) for w in x_cv])
#print("Accuracy on CV set: {0}".format(acc_cv))