def test_split_random_punct(self):
for i in range(1, 100):
r = random.randint(1, i)
word = u''
for j in range(1, r):
p = random.randint(0, len(string.punctuation) - 1)
word += string.punctuation[p]
result = split([word])
self.assertEqual(list(result), list(word))
def test_split_random_punct(self):
for i in range(1, 100):
r = random.randint(1, i)
word = u''
for j in range(1, r):
p = random.randint(0, len(string.punctuation) - 1)
word += string.punctuation[p]
result = split([word])
expected = list(word)
self.assertEqual(list(result), expected)
def test_split(self):
"""Split tokens into terms using the following rules:
0. All digits are discarded
1. A sequence beginning with an lc letter must be followed by lc letters
2. A sequence beginning with an uc letter can be followed by either:
a. One or more uc letters
b. One or more lc letters
"""
cases = dict({
'camelCase': ('camel', 'Case'),
'CamelCase': ('Camel', 'Case'),
'camel2case': ('camel', '2', 'case'),
'camel2Case': ('camel', '2', 'Case'),
'word': ('word', ),
'HTML': ('HTML', ),
'readXML': ('read', 'XML'),
'XMLRead': ('XML', 'Read'),
'firstMIDDLELast': ('first', 'MIDDLE', 'Last'),
'CFile': ('C', 'File'),
'Word2Word34': ('Word', '2', 'Word', '34'),
'WORD123Word': ('WORD', '123', 'Word'),
'c_amelCase': ('c', '_', 'amel', 'Case'),
'CamelC_ase': ('Camel', 'C', '_', 'ase'),
'camel2_case': ('camel', '2', '_', 'case'),
'camel_2Case': ('camel', '_', '2', 'Case'),
'word': ('word', ),
'HTML': ('HTML', ),
'read_XML': ('read', '_', 'XML'),
'XML_Read': ('XML', '_', 'Read'),
'firstM_IDDL_ELast': ('first', 'M', '_', 'IDDL', '_', 'E', 'Last'),
'the_CFile': ('the', '_', 'C', 'File'),
'Word_2_Word3_4': ('Word', '_', '2', '_', 'Word', '3', '_', '4'),
'WO_RD123W_or_d':
('WO', '_', 'RD', '123', 'W', '_', 'or', '_', 'd'),
'hypen-ation': ('hypen', '-', 'ation'),
'*****@*****.**': ('email', '@', 'address', '.', 'com'),
'/*comment*/': ('/', '*', 'comment', '*', '/'),
'word1': ('word', '1'),
'Word1': ('Word', '1'),
'f1': ('f', '1'),
'1ms': ('1', 'ms'),
'F1': ('F', '1'),
'WORD_THING': ('WORD', '_', 'THING'),
'@': ('@', ),
'WORD_THING_ONE': ('WORD', '_', 'THING', '_', 'ONE'),
'wordThing_one': ('word', 'Thing', '_', 'one'),
'_w': ('_', 'w'),
'_wt': ('_', 'wt'),
'_wT': ('_', 'w', 'T'),
'_WT': ('_', 'WT'),
'_Wt': ('_', 'Wt'),
'wt_': ('wt', '_'),
'<5>': ('<', '5', '>'),
'==': ('=', '='),
'x=5;': ('x', '=', '5', ';'),
'2.0': ('2', '.', '0'),
'2,0': ('2', ',', '0'),
'//test': ('/', '/', 'test'),
'Boolean.FALSE': ('Boolean', '.', 'FALSE'),
'word': ('word', ),
'word.': ('word', '.'),
'.word.': ('.', 'word', '.'),
'.word': ('.', 'word'),
'WordThing.': ('Word', 'Thing', '.'),
'WordThing.FLAG': ('Word', 'Thing', '.', 'FLAG'),
'WordThing.cmd': ('Word', 'Thing', '.', 'cmd'),
'WordThing.cmdDo': ('Word', 'Thing', '.', 'cmd', 'Do'),
'System.out.println': ('System', '.', 'out', '.', 'println'),
'System.out.println();':
('System', '.', 'out', '.', 'println', '(', ')', ';'),
'x++': ('x', '+', '+'),
'++x': ('+', '+', 'x'),
"n't": ('n', "'", 't'),
u"test💩word": ('test', u'💩', 'word'),
u'Erwin_Schrödinger': ('Erwin', '_', u'Schrödinger')
})
for term, expected in cases.items():
result = split([term])
self.assertEqual(tuple(result), expected)
"""
def test_split_creates_generator(self):
""" Split tokens creates a generator """
result = split('butts')
self.assertIsInstance(result, type(x for x in list()))
def test_split(self):
"""Split tokens into terms using the following rules:
0. All digits are discarded
1. A sequence beginning with an lc letter must be followed by lc letters
2. A sequence beginning with an uc letter can be followed by either:
a. One or more uc letters
b. One or more lc letters
"""
cases = dict({
'camelCase': ('camel', 'Case'),
'CamelCase': ('Camel', 'Case'),
'camel2case': ('camel', '2', 'case'),
'camel2Case': ('camel', '2', 'Case'),
'word': ('word', ),
'HTML': ('HTML', ),
'readXML': ('read', 'XML'),
'XMLRead': ('XML', 'Read'),
'firstMIDDLELast': ('first', 'MIDDLE', 'Last'),
'CFile': ('C', 'File'),
'Word2Word34': ('Word', '2', 'Word', '34'),
'WORD123Word': ('WORD', '123', 'Word'),
'c_amelCase': ('c', '_', 'amel', 'Case'),
'CamelC_ase': ('Camel', 'C', '_', 'ase'),
'camel2_case': ('camel', '2', '_', 'case'),
'camel_2Case': ('camel', '_', '2', 'Case'),
'word': ('word', ),
'HTML': ('HTML', ),
'read_XML': ('read', '_', 'XML'),
'XML_Read': ('XML', '_', 'Read'),
'firstM_IDDL_ELast': ('first', 'M', '_', 'IDDL', '_', 'E', 'Last'),
'the_CFile': ('the', '_', 'C', 'File'),
'Word_2_Word3_4': ('Word', '_', '2', '_', 'Word', '3', '_', '4'),
'WO_RD123W_or_d': ('WO', '_', 'RD', '123', 'W', '_', 'or', '_', 'd'),
'hypen-ation': ('hypen', '-', 'ation'),
'*****@*****.**': ('email', '@', 'address', '.', 'com'),
'/*comment*/': ('/', '*', 'comment', '*', '/'),
'word1': ('word', '1'),
'Word1': ('Word', '1'),
'f1': ('f', '1'),
'1ms': ('1', 'ms'),
'F1': ('F', '1'),
'WORD_THING': ('WORD', '_', 'THING'),
'@': ('@',),
'WORD_THING_ONE': ('WORD', '_', 'THING', '_', 'ONE'),
'wordThing_one': ('word', 'Thing', '_', 'one'),
'_w': ('_', 'w'),
'_wt': ('_', 'wt'),
'_wT': ('_', 'w', 'T'),
'_WT': ('_', 'WT'),
'_Wt': ('_', 'Wt'),
'wt_': ('wt', '_'),
'<5>': ('<', '5', '>'),
'==': ('=', '='),
'x=5;': ('x', '=', '5', ';'),
'2.0': ('2', '.', '0'),
'2,0': ('2', ',', '0'),
'//test': ('/', '/', 'test'),
'Boolean.FALSE': ('Boolean', '.', 'FALSE'),
'word': ('word', ),
'word.': ('word', '.'),
'.word.': ('.', 'word', '.'),
'.word': ('.', 'word'),
'WordThing.': ('Word', 'Thing', '.'),
'WordThing.FLAG': ('Word', 'Thing', '.', 'FLAG'),
'WordThing.cmd': ('Word', 'Thing', '.', 'cmd'),
'WordThing.cmdDo': ('Word', 'Thing', '.', 'cmd', 'Do'),
'System.out.println': ('System', '.', 'out', '.', 'println'),
'System.out.println();': ('System', '.', 'out', '.', 'println', '(', ')', ';'),
'x++': ('x', '+', '+'),
'++x': ('+', '+', 'x'),
"n't": ('n', "'", 't'),
u"test💩word": ('test', u'💩', 'word'),
u'Erwin_Schrödinger': ('Erwin', '_', u'Schrödinger')
})
for term, expected in cases.items():
result = split([term])
self.assertEqual(tuple(result), expected)
"""
def test_split_creates_generator(self):
""" Split tokens creates a generator """
result = split('butts')
self.assertIsInstance(result, type(x for x in list()))
import numpy as np
from src.activation_functions import ReLU, Softmax, Sin, Cos
from src.evaluation import plot_loss_and_accuracy, accuracy
from src.loss_functions import SquaredLoss, CrossEntropy
from src.neural_net.layers import InputLayer, Layer
from src.neural_net.network import NeuralNetwork
from src.preprocessing import to_categorical, split
from keras.datasets import mnist
(X, Y), (Xtest, Ytest) = mnist.load_data()
# Reshape & Normalize
dimensions = X.shape[1] * X.shape[2] # i.e. flattened
X = X.reshape((X.shape[0], dimensions)) / 255.0
Xtest = Xtest.reshape((Xtest.shape[0], dimensions)) / 255.0
(X, Y), (Xval, Yval) = split(X, Y, train_frac=0.9)
Y = to_categorical(Y)
hidden_layer_act = ReLU()
layers = [
InputLayer(X.shape[1]),
Layer(25, hidden_layer_act),
Layer(25, hidden_layer_act),
Layer(10, Softmax())
]
nn = NeuralNetwork(layers)
nn.compile(loss_function=SquaredLoss(), metric=accuracy)
history = nn.fit(X,
Y,
