# -*- coding: UTF-8 -*-
import numpy as np
from ocr import OCRNeuralNetwork
from sklearn.cross_validation import train_test_split
def test(data_matrix, data_labels, test_indices, nn):
correct_guess_count = 0
for i in test_indices:
test = data_matrix[i]
prediction = nn.predict(test)
if data_labels[i] == prediction:
correct_guess_count += 1
return correct_guess_count / float(len(test_indices))
data_matrix = np.loadtxt(open('../data/simple_ocr/data.csv', 'rb'), delimiter = ',').tolist()
data_labels = np.loadtxt(open('../data/simple_ocr/dataLabels.csv', 'rb')).tolist()
# Create training and testing sets.
train_indices, test_indices = train_test_split(list(range(5000)))
print "PERFORMANCE"
print "-----------"
for i in xrange(5, 50, 5):
nn = OCRNeuralNetwork(i, data_matrix, data_labels, train_indices, False)
performance = str(test(data_matrix, data_labels, test_indices, nn))
print "{i} Hidden Nodes: {val}".format(i=i, val=performance)
HOST_NAME = 'localhost'
PORT_NUMBER = 8000
HIDDEN_NODE_COUNT = 15
# Load data samples and labels into matrix
data_matrix = np.loadtxt(open('data.csv', 'rb'), delimiter=',')
data_labels = np.loadtxt(open('dataLabels.csv', 'rb'))
# Convert from numpy ndarrays to python lists
data_matrix = data_matrix.tolist()
data_labels = data_labels.tolist()
# If a neural network file does not exist, train it using all 5000 existing data samples.
# Based on data collected from neural_network_design.py, 15 is the optimal number
# for hidden nodes
nn = OCRNeuralNetwork(HIDDEN_NODE_COUNT, data_matrix, data_labels,
random.sample(range(5000), 5000))
# The origin code did not shuffle the data, which makes predict output 9 every time!
class JSONHandler(BaseHTTPServer.BaseHTTPRequestHandler):
def do_POST(s):
response_code = 200
response = ""
var_len = int(s.headers.get('Content-Length'))
content = s.rfile.read(var_len)
payload = json.loads(content)
if payload.get('train'):
nn.train(payload['trainArray'])
nn.save()
HIDDEN_NODE_COUNT = 20
# 加载数据集
data_matrix = np.loadtxt(open('data.csv', 'rb'), delimiter=',')
data_labels = np.loadtxt(open('dataLabels.csv', 'rb'))
# 转换成list类型
data_matrix = data_matrix.tolist()
data_labels = data_labels.tolist()
# 数据集一共5000个数据，train_indice存储用来训练的数据的序号
train_indice = list(range(5000))
# 打乱训练顺序
random.shuffle(train_indice)
nn = OCRNeuralNetwork(HIDDEN_NODE_COUNT, data_matrix, data_labels,
train_indice)
class JSONHandler(SimpleHTTPRequestHandler):
"""处理接收到的POST请求"""
def do_POST(self):
response_code = 200
response = ""
var_len = int(self.headers.get('Content-Length'))
content = self.rfile.read(var_len)
payload = json.loads(content)
# 如果是训练请求，训练然后保存训练完的神经网络
if payload.get('train'):
nn.train(payload['trainArray'])
nn.save()
HOST_NAME = 'localhost'
PORT_NUMBER = 8000
HIDDEN_NODE_COUNT = 15
# Load data samples and labels into matrix
data_matrix = np.loadtxt(open('data.csv', 'rb'), delimiter=',')
data_labels = np.loadtxt(open('dataLabels.csv', 'rb'))
# Convert from numpy ndarrays to python lists
data_matrix = data_matrix.tolist()
data_labels = data_labels.tolist()
# If a neural network file does not exist, train it using all 5000 existing data samples.
# Based on data collected from neural_network_design.py, 15 is the optimal number
# for hidden nodes
nn = OCRNeuralNetwork(HIDDEN_NODE_COUNT, data_matrix, data_labels,
list(range(5000)))
class JSONHandler(BaseHTTPServer.BaseHTTPRequestHandler):
def do_POST(s):
response_code = 200
response = ""
var_len = int(s.headers.get('Content-Length'))
content = s.rfile.read(var_len)
payload = json.loads(content)
if payload.get('train'):
nn.train(payload['trainArray'])
nn.save()
elif payload.get('predict'):
try:
cnt += 1
return cnt / float(len(test_indices))
# Load data samples and labels into matrix
data_matrix = np.loadtxt(open('mydata.csv', 'rb'), delimiter=',').tolist()
data_labels = np.loadtxt(open('mydataLabels.csv', 'rb')).tolist()
# Create training and testing sets.
train_indices, test_indices = train_test_split(list(range(len(data_matrix))))
print "PERFORMANCE"
print "-----------"
maxi, maxj = 10, 1
maxnn = OCRNeuralNetwork(400, 10, 10, data_matrix, data_labels, train_indices,
1)
maxp = test(data_matrix, data_labels, test_indices, maxnn)
for i in xrange(10, 50):
for j in xrange(1, 10):
nn = OCRNeuralNetwork(400, i, 10, data_matrix, data_labels,
train_indices, j)
p = test(data_matrix, data_labels, test_indices, nn)
if p > maxp:
maxi, maxj, maxp = i, j, p
performance = str(p)
print "{i} Hidden Nodes, {j} trainings: {val}".format(i=i,
j=j,
val=performance)
print 'max:', maxi, 'Hidden Nodes,', maxj, 'trainings:', str(maxp)
PORT_NUMBER = 8000
INPUT_NODE_COUNT = 400
HIDDEN_NODE_COUNT = 49
NUM_OF_TRAINGING = 8
OUTPUT_NODE_COUNT = 10
data_matrix = np.loadtxt(open('mydata.csv', 'rb'), delimiter=',')
data_labels = np.loadtxt(open('mydataLabels.csv', 'rb'))
data_matrix = data_matrix.tolist()
data_labels = data_labels.tolist()
print 'begin train'
nn = OCRNeuralNetwork(INPUT_NODE_COUNT, HIDDEN_NODE_COUNT,
OUTPUT_NODE_COUNT, data_matrix, data_labels,
list(range(len(data_matrix))), NUM_OF_TRAINGING)
print 'end train'
nn.save()
class JSONHandler(BaseHTTPServer.BaseHTTPRequestHandler):
def do_POST(s):
response_code = 200
response = ""
var_len = int(s.headers.get('Content-Length'))
content = s.rfile.read(var_len)
payload = json.loads(content)
if payload.get('train'):
TrainData = namedtuple('TrainData', ['y0', 'label'])