# encoding=utf8
import os
import random
import numpy as np
import FukuML.Utility as utility
import FukuML.MLBase as ml
random.seed(1)
class NeuralNetwork(ml.Learner):
def __init__(self):
"""init"""
self.status = 'empty'
self.train_X = []
self.train_Y = []
self.W = []
self.data_num = 0
self.data_demension = 0
self.test_X = []
self.test_Y = []
self.feature_transform_mode = ''
self.feature_transform_degree = 1
self.network_structure = []
self.w_range_high = 0.1
self.w_range_low = -0.1
self.feed_mode = 'stochastic'
self.step_eta = 0.01
self.updates = 50000
def load_train_data(self, input_data_file=''):
self.status = 'load_train_data'
if (input_data_file == ''):
input_data_file = os.path.normpath(os.path.join(os.path.join(os.getcwd(), os.path.dirname(__file__)), "dataset/neural_network_train.dat"))
else:
if (os.path.isfile(input_data_file) is not True):
print("Please make sure input_data_file path is correct.")
return self.train_X, self.train_Y
self.train_X, self.train_Y = utility.DatasetLoader.load(input_data_file)
return self.train_X, self.train_Y
def load_test_data(self, input_data_file=''):
if (input_data_file == ''):
input_data_file = os.path.normpath(os.path.join(os.path.join(os.getcwd(), os.path.dirname(__file__)), "dataset/neural_network_test.dat"))
else:
if (os.path.isfile(input_data_file) is not True):
print("Please make sure input_data_file path is correct.")
return self.test_X, self.test_Y
self.test_X, self.test_Y = utility.DatasetLoader.load(input_data_file)
if (self.feature_transform_mode == 'polynomial') or (self.feature_transform_mode == 'legendre'):
self.test_X = self.test_X[:, 1:]
self.test_X = utility.DatasetLoader.feature_transform(
self.test_X,
self.feature_transform_mode,
self.feature_transform_degree
)
return self.test_X, self.test_Y
def set_param(self, network_structure=[8, 3, 1], w_range_high=0.1, w_range_low=-0.1, feed_mode='stochastic', step_eta=0.01, updates=50000):
self.network_structure = network_structure
self.w_range_high = w_range_high
self.w_range_low = w_range_low
self.feed_mode = feed_mode
self.step_eta = step_eta
self.updates = updates
return self.network_structure, self.w_range_high, self.w_range_low, self.feed_mode, self.step_eta, self.updates
def init_W(self):
if (self.status != 'load_train_data') and (self.status != 'train'):
print("Please load train data first.")
return self.W
self.status = 'init'
self.data_num = len(self.train_Y)
self.data_demension = len(self.train_X[0])
self.network_structure.insert(0, self.data_demension - 1)
for i in range(1, len(self.network_structure)):
this_layer_w = np.random.uniform(self.w_range_low, self.w_range_high, (self.network_structure[i - 1] + 1, self.network_structure[i]))
self.W.append(this_layer_w)
return self.W
def score_function(self, x, W):
y_predict = x[1:]
for i in range(0, len(W), 1):
y_predict = np.tanh(np.dot(np.hstack((1, y_predict)), W[i]))
score = y_predict[0]
return score
def error_function(self, y_prediction, y_truth):
error = (y_prediction - y_truth) ** 2
return error
def calculate_avg_error(self, X, Y, W):
return super(NeuralNetwork, self).calculate_avg_error(X, Y, W)
def calculate_test_data_avg_error(self):
return super(NeuralNetwork, self).calculate_test_data_avg_error()
def tanh_prime(self, s):
tanh_prime_output = np.zeros(s.shape)
for i in range(s.shape[0]):
tanh_prime_output[i] = 4.0 / (np.exp(2 * s[i]) + np.exp(-2 * s[i]) + 2)
return tanh_prime_output
def forward_process(self, x, y, W):
forward_output = []
pre_x = x
for i in range(len(W)):
pre_x = np.tanh(np.dot(pre_x, W[i]))
forward_output.append(pre_x)
pre_x = np.hstack((1, pre_x))
return forward_output
def backward_process(self, x, y, W, neuron_output):
backward_output = []
layer_num = len(neuron_output)
score = np.dot(np.hstack((1, neuron_output[layer_num - 2])), W[layer_num - 1])
error_gradient = np.array([-2 * (y - neuron_output[layer_num - 1][0]) * self.tanh_prime(score)])
# error_gradient = np.array([np.sum(-2 * (y - score) * np.hstack((1, neuron_output[layer_num-2])))])
backward_output.insert(0, error_gradient)
# Hidden layer
for i in range(layer_num - 2, -1, -1):
if i == 0:
score = np.dot(x, W[i])
else:
score = np.dot(np.hstack((1, neuron_output[i - 1])), W[i])
error_gradient = np.dot(error_gradient, W[i + 1][1:].transpose()) * self.tanh_prime(score)
backward_output.insert(0, error_gradient)
return backward_output
def update_W_with_gradient_decent(self, x, neuron_output, error_gradient):
w_output = []
layer_num = len(self.W)
w_output.append(self.W[0] - self.step_eta * np.array([x]).transpose() * error_gradient[0])
for i in range(1, layer_num, 1):
w_output.append(self.W[i] - self.step_eta * np.array([np.hstack((1, neuron_output[i - 1]))]).transpose() * error_gradient[i])
return w_output
def train(self):
if (self.status != 'init'):
print("Please load train data and init W first.")
return self.W
self.status = 'train'
for i in range(0, self.updates):
stochastic_i = random.randint(0, self.data_num - 1)
x = self.train_X[stochastic_i]
y = self.train_Y[stochastic_i]
neuron_output = self.forward_process(x, y, self.W)
error_gradient = self.backward_process(x, y, self.W, neuron_output)
self.W = self.update_W_with_gradient_decent(x, neuron_output, error_gradient)
return self.W
def prediction(self, input_data='', mode='test_data'):
return super(NeuralNetwork, self).prediction(input_data, mode)
class BinaryClassifier(NeuralNetwork):
def __init__(self):
"""init"""
self.status = 'empty'
self.train_X = []
self.train_Y = []
self.W = []
self.data_num = 0
self.data_demension = 0
self.test_X = []
self.test_Y = []
self.feature_transform_mode = ''
self.feature_transform_degree = 1
self.network_structure = []
self.w_range_high = 0.1
self.w_range_low = -0.1
self.feed_mode = 'stochastic'
self.step_eta = 0.01
self.updates = 50000
def load_train_data(self, input_data_file=''):
return super(BinaryClassifier, self).load_train_data(input_data_file)
def load_test_data(self, input_data_file=''):
return super(BinaryClassifier, self).load_test_data(input_data_file)
def set_param(self, network_structure=[8, 3, 1], w_range_high=0.1, w_range_low=-0.1, feed_mode='stochastic', step_eta=0.01, updates=50000):
return super(BinaryClassifier, self).set_param(network_structure, w_range_high, w_range_low, feed_mode, step_eta, updates)
def init_W(self):
return super(BinaryClassifier, self).init_W()
def score_function(self, x, W):
score = super(BinaryClassifier, self).score_function(x, W)
score = np.sign(score)
return score
def error_function(self, y_prediction, y_truth):
if y_prediction != y_truth:
return 1
else:
return 0
def calculate_avg_error(self, X, Y, W):
return super(BinaryClassifier, self).calculate_avg_error(X, Y, W)
def calculate_test_data_avg_error(self):
return super(BinaryClassifier, self).calculate_test_data_avg_error()
def tanh_prime(self, s):
return super(BinaryClassifier, self).tanh_prime(s)
def forward_process(self, x, y, W):
return super(BinaryClassifier, self).forward_process(x, y, W)
def backward_process(self, x, y, W, neuron_output):
return super(BinaryClassifier, self).backward_process(x, y, W, neuron_output)
def update_W_with_gradient_decent(self, x, neuron_output, error_gradient):
return super(BinaryClassifier, self).update_W_with_gradient_decent(x, neuron_output, error_gradient)
def train(self):
return super(BinaryClassifier, self).train()
def prediction(self, input_data='', mode='test_data'):
return super(BinaryClassifier, self).prediction(input_data, mode)
