ml-finance-python
python scripts for finance machine learning
git clone https://9o.is/git/ml-finance-python.git
convolutional_neural_network.py
(2904B)
1
2 from __future__ import print_function
3 from sklearn import datasets
4 import matplotlib.pyplot as plt
5 import math
6 import numpy as np
7
8 # Import helper functions
9 from mlfromscratch.deep_learning import NeuralNetwork
10 from mlfromscratch.utils import train_test_split, to_categorical, normalize
11 from mlfromscratch.utils import get_random_subsets, shuffle_data, Plot
12 from mlfromscratch.utils.data_operation import accuracy_score
13 from mlfromscratch.deep_learning.optimizers import StochasticGradientDescent, Adam, RMSprop, Adagrad, Adadelta
14 from mlfromscratch.deep_learning.loss_functions import CrossEntropy
15 from mlfromscratch.utils.misc import bar_widgets
16 from mlfromscratch.deep_learning.layers import Dense, Dropout, Conv2D, Flatten, Activation, MaxPooling2D
17 from mlfromscratch.deep_learning.layers import AveragePooling2D, ZeroPadding2D, BatchNormalization, RNN
18
19
20
21 def main():
22
23 #----------
24 # Conv Net
25 #----------
26
27 optimizer = Adam()
28
29 data = datasets.load_digits()
30 X = data.data
31 y = data.target
32
33 # Convert to one-hot encoding
34 y = to_categorical(y.astype("int"))
35
36 X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4, seed=1)
37
38 # Reshape X to (n_samples, channels, height, width)
39 X_train = X_train.reshape((-1,1,8,8))
40 X_test = X_test.reshape((-1,1,8,8))
41
42 clf = NeuralNetwork(optimizer=optimizer,
43 loss=CrossEntropy,
44 validation_data=(X_test, y_test))
45
46 clf.add(Conv2D(n_filters=16, filter_shape=(3,3), stride=1, input_shape=(1,8,8), padding='same'))
47 clf.add(Activation('relu'))
48 clf.add(Dropout(0.25))
49 clf.add(BatchNormalization())
50 clf.add(Conv2D(n_filters=32, filter_shape=(3,3), stride=1, padding='same'))
51 clf.add(Activation('relu'))
52 clf.add(Dropout(0.25))
53 clf.add(BatchNormalization())
54 clf.add(Flatten())
55 clf.add(Dense(256))
56 clf.add(Activation('relu'))
57 clf.add(Dropout(0.4))
58 clf.add(BatchNormalization())
59 clf.add(Dense(10))
60 clf.add(Activation('softmax'))
61
62 print ()
63 clf.summary(name="ConvNet")
64
65 train_err, val_err = clf.fit(X_train, y_train, n_epochs=50, batch_size=256)
66
67 # Training and validation error plot
68 n = len(train_err)
69 training, = plt.plot(range(n), train_err, label="Training Error")
70 validation, = plt.plot(range(n), val_err, label="Validation Error")
71 plt.legend(handles=[training, validation])
72 plt.title("Error Plot")
73 plt.ylabel('Error')
74 plt.xlabel('Iterations')
75 plt.show()
76
77 _, accuracy = clf.test_on_batch(X_test, y_test)
78 print ("Accuracy:", accuracy)
79
80
81 y_pred = np.argmax(clf.predict(X_test), axis=1)
82 X_test = X_test.reshape(-1, 8*8)
83 # Reduce dimension to 2D using PCA and plot the results
84 Plot().plot_in_2d(X_test, y_pred, title="Convolutional Neural Network", accuracy=accuracy, legend_labels=range(10))
85
86 if __name__ == "__main__":
87 main()