def aa_substring(string=None):
print('Testing DFA, Detect if \'aa\' is a substring of w')
sigma = ['a', 'b']
Q = ['q1', 'q2', 'q3']
init_state = 'q1'
delta = dict()
delta[('q1', 'a')] = 'q2'
delta[('q1', 'b')] = 'q1'
delta[('q2', 'a')] = 'q3'
delta[('q2', 'b')] = 'q1'
delta[('q3', 'a')] = 'q3'
delta[('q3', 'b')] = 'q3'
final = ['q3']
my_auto = dfa.DFA(sigma, Q, init_state, delta, final)
if string is None:
inputs = ['aabbbb', 'bbbbbbb', 'bababaabbba', 'abababababa']
for inp in inputs:
if my_auto.consume_input(inp):
print('{} was accepted'.format(inp))
else:
print('{} was not accepted'.format(inp))
else:
if my_auto.consume_input(string):
print('{} was accepted'.format(string))
else:
print('{} was not accepted'.format(string))
def toDFA(self):
d = dfa.DFA(self.sigma)
q = queue.Queue()
mp = dict()
f = frozenset(self.init_state)
q.put(f)
cnt = 0
mp[f] = cnt
d.add_node(cnt)
cnt += 1
d.set_init_state(0)
while not q.empty():
u = q.get()
# print(u)
for w in self.sigma:
v = self.closure(u, w)
if len(v) == 0:
continue
v = frozenset(v)
print(u, v)
if not v in mp:
mp[v] = cnt
d.add_node(cnt)
for flabel in self.final_state:
if flabel in v:
d.final_state.append(cnt)
break
cnt += 1
q.put(v)
d.add_edge(mp[u], mp[v], w)
return d
def __init__(self, ip, name, position, sensors):
""" -> String -> "in" / "out" -> [String] -> Station() """
self.ip = ip
self.name = name
self.sensors = Sensors(sensors)
self.position = position
self.dfa = df.DFA()
def multof3(string=None):
print('Testing DFA, Detect if number of 0\'s is a multiple of 3')
sigma = ['1', '0']
Q = ['r1', 'r2', 'r3']
init_state = 'r1'
delta = dict()
delta[('r1', '0')] = 'r2'
delta[('r1', '1')] = 'r1'
delta[('r2', '0')] = 'r3'
delta[('r2', '1')] = 'r2'
delta[('r3', '0')] = 'r1'
delta[('r3', '1')] = 'r3'
final = ['r1']
a = '{} was accepted by the DFA'
n = '{} was not accepted by the DFA'
my_dfa = dfa.DFA(sigma, Q, init_state, delta, final)
if string is None:
inputs = ['100100', '1111', '10111010', '0000', '000']
for inp in inputs:
if my_dfa.consume_input(inp):
print('{} was accepted'.format(inp))
else:
print('{} was not accepted'.format(inp))
else:
if my_dfa.consume_input(string):
print('{} was accepted'.format(string))
else:
print('{} was not accepted'.format(string))
print('\n')
def multof2(string=None):
print('Testing DFA, Detect if number of 0\'s is even (multiple of 2)')
sigma = ['1', '0']
Q = ['q1', 'q2']
init_state = 'q1'
delta = dict()
delta[('q1', '1')] = 'q1'
delta[('q1', '0')] = 'q2'
delta[('q2', '0')] = 'q1'
delta[('q2', '1')] = 'q2'
final = ['q1']
a = '{} was accepted by the DFA'
n = '{} was not accepted by the DFA'
my_dfa = dfa.DFA(sigma, Q, init_state, delta, final)
if string is None:
inputs = ['100100', '1111', '10111010', '0000', '000']
results = {}
for inp in inputs:
try:
results[inp] = my_dfa.consume_input(inp)
if results[inp]:
print(a.format(inp))
else:
print(n.format(inp))
except util.stateViolation as e:
print(e)
else:
if my_dfa.consume_input(string):
print('{} was accepted'.format(string))
else:
print('{} was not accepted'.format(string))
print('\n')
def determinize(self):
# based on:
# Subset Construction presented by Ullman (Automata course)
# extended with closure operation to work with epsilon transitions
# if not self.is_complete():
# self.set_error_state()
closure = None
if not self.is_epsilon_free():
closure = self.closure()
dfa_accept = set()
dfa_reject = set()
if closure is None:
start = frozenset(self.start)
else:
start = frozenset({ x for s in self.start \
for x in closure[s] })
dfa_states = {start}
tr = collections.defaultdict(lambda: None)
queue = collections.deque([start])
if start & self.accept:
dfa_accept.add(start)
elif start & self.reject:
dfa_reject.add(start)
while queue:
curr_state = queue.popleft()
for a in self.alphabet:
next_direct = { x for s in curr_state \
for x in self.delta(s, a) }
if closure is None:
next_state = frozenset(next_direct)
else:
next_state = frozenset({ x for s in next_direct \
for x in closure[s] })
if next_state:
if next_state not in dfa_states:
dfa_states.add(next_state)
queue.append(next_state)
tr[curr_state, a] = next_state
if next_state & self.accept:
dfa_accept.add(next_state)
elif next_state & self.reject:
dfa_reject.add(next_state)
# dfa_delta = lambda s,a: tr.get((s, a), None)
dfa_delta = lambda q, a: tr[q, a]
return dfa.DFA(self.alphabet, dfa_states, start, dfa_accept,
dfa_reject, dfa_delta, tr)
def setUp(self):
self.alphabets = set(['0', '1'])
self.states = set(['a', 'b'])
self.initialState = 'a'
self.endState = set(list('b'))
self.transitionRule = {
('a', '0'): 'b',
('a', '1'): 'a',
('b', '0'): 'b',
('b', '1'): 'a'
}
self.transitionRuleIncomplete = {
('a', '0'): 'b',
('a', '1'): 'a',
('b', '0'): 'b',
}
self.dfaTrue = dfa.DFA(self.alphabets, self.states, self.initialState,
self.endState, self.transitionRule)
self.dfaFalse = dfa.DFA(self.alphabets, self.states, self.initialState,
self.endState, self.transitionRuleIncomplete)
def train(total_loss, global_step):
"""Train CIFAR-10 model.
Create an optimizer and apply to all trainable variables. Add moving
average for all trainable variables.
Args:
total_loss: Total loss from loss().
global_step: Integer Variable counting the number of training steps
processed.
Returns:
train_op: op for training.
"""
# Variables that affect learning rate.
num_batches_per_epoch = NUM_EXAMPLES_PER_EPOCH_FOR_TRAIN / FLAGS.batch_size
# Decay the learning rate exponentially based on the number of steps.
lr = INITIAL_LEARNING_RATE
tf.summary.scalar('learning_rate', lr)
# Generate moving averages of all losses and associated summaries.
#loss_averages_op = _add_loss_summaries(total_loss)
losses = tf.get_collection('losses')
for l in losses + [total_loss]:
tf.summary.scalar(l.op.name + ' (raw)', l)
# Compute gradients.
#with tf.control_dependencies([loss_averages_op]):
opt = dfa.DFA(lr, stddev=1.0) # Direct Feedback Alignment optimizer
grads = opt.compute_gradients(total_loss) # Call custom compute_gradients method
maxnorm = [(tf.clip_by_norm(grad, 4), var) for grad, var in grads]
# Apply gradients.
apply_gradient_op = opt.apply_gradients(maxnorm, global_step=global_step) # Used _apply_dense function from tensorflow's stochastic gradient descent, see dfa.py
# Add histograms for trainable variables.
for var in tf.trainable_variables():
tf.summary.histogram(var.op.name, var)
# Add histograms for gradients.
for grad, var in grads:
if grad is not None:
tf.summary.histogram(var.op.name + '/gradients', grad)
# Track the moving averages of all trainable variables.
variable_averages = tf.train.ExponentialMovingAverage(
MOVING_AVERAGE_DECAY, global_step)
variables_averages_op = variable_averages.apply(tf.trainable_variables())
with tf.control_dependencies([apply_gradient_op, variables_averages_op]):
train_op = tf.no_op(name='train')
return train_op
def get(self, query):
tokens = process_tokens(query)
dfa_ = dfa.DFA()
for token in tokens:
print token
print dfa_.state
dfa_.next(token)
print dfa_.state
ret = {
'state': dfa_.state.get_name(), # For testing
'suggestions': dfa_.state.get_suggestions()
}
return jsonify(ret), 200
def get(self, query):
calls = []
tokens = process_tokens(query)
dfa_ = dfa.DFA()
for token in tokens:
dfa_.next(token)
for call in dfa_.get_calls():
print call
for call in dfa_.translate_calls():
calls.append(create_api_call(call[0], call[1], call[2]))
if len(calls) != 0:
return jsonify({'apiCalls': calls}), 200
return jsonify({'error': 'Please try again'}), 400
def toDFA(self):
exp = self.exp
operators = ('*', '+', '|')
alphabet = set(exp) - set(operators)
alphabet = list(alphabet)
count = 0
key = str(count)
# mapping from new keys to the aggregated states
states = dict()
states[key] = self.findEpsilon([self.startState])
# record relations between keys
paths = dict()
# store the keys to be processed
queue = deque(list())
queue.append(key)
while len(queue):
newkey = queue.popleft()
for c in alphabet:
newstate = self.findCondition(states[newkey], c)
newstate = self.findEpsilon(newstate)
key = self.getKey(states, newstate)
if key == 0:
count += 1
key = str(count)
states[key] = newstate
queue.append(key)
if paths.has_key(newkey) == False:
paths[newkey] = dict()
paths[newkey][c] = key
sStates = list()
fStates = list()
for key in states.keys():
if self.startState in states[key]:
sStates.append(key)
if self.finalState in states[key]:
fStates.append(key)
# print states
states = states.keys()
d = dfa.DFA()
d.setStates(states, paths, sStates, fStates, alphabet)
return d
def main():
try:
fileName = sys.argv[1]
except IndexError:
fileName = input('Enter the program Name: ')
userParser = parser.Parser(fileName)
userParser.format()
try:
userDfa = dfa.DFA(userParser.dfaValues[0], userParser.dfaValues[1],
userParser.dfaValues[2], userParser.dfaValues[3],
userParser.dfaValues[4])
except:
exit('Syntax Error Ocuured')
print('Your DFA has been loaded to the shell.')
string = takeInput()
while len(string) >= 0:
try:
print(userDfa.start(string))
except exceptionCollection.StringError:
print('Please enter the valid string')
string = takeInput()
def to_dfa(self):
'''
将NFA转化为DFA
:return: dfa.DFA对象
设置DFA的非终结符为自然数
DFA的终结符号集与NFA相同
'''
DFA_index = 0 # 设置DFA的非终结符为自然数
# DFA_VN = dict() # DFA的非终结符号集
DFA_f = dict() # DFA的映射函数
DFA_S = self.epsilon_closure(self.S) # DFA的初始状态
DFA_T = [] # DFA的终止状态集
self.C.append(DFA_S)
while True:
T = self.C[DFA_index]
# DFA_VN[DFA_index] = T
if not T.isdisjoint(self.T): # 如果T包含NFA终止状态集的元素，则T为DFA的终止状态
DFA_T.append(DFA_index)
DFA_f[DFA_index] = dict()
for ch in self.VT:
U = self.epsilon_closure(self.move(T, ch))
if len(U) == 0: # 如果为空集，处理下一个终结符
continue
if U not in self.C:
self.C.append(U)
self.set_dfa_f(
DFA_index, ch, U,
DFA_f) # 利用当前self.C的数据，将NFA非终止状态的子集映射到表示非终止状态的自然数
if DFA_index + 1 == len(self.C): # 若没有新的DFA状态可扩展，则转化过程结束
break
DFA_index += 1 # 准备进行下一个DFA状态的转化
new_dfa = dfa.DFA([i for i in range(len(self.C))], self.VT, DFA_f, 0,
DFA_T) # 生成的新DFA
return new_dfa
def import_rules(path):
''' import DFAs from rule file.
:parameter
path : string
path of the rule file,
:return
set of the DFAs imported.
'''
fr = open(path, 'r')
data = json.load(fr)
rules = dict()
# order is a list which define the order of the rules
order = data['names']
rules['order'] = order
for name in order:
d = dfa.DFA()
d.decode(data[name])
rules[name] = d
return rules
def create_new():
structure_type = ''
while True:
structure_type = input('Create: (D)FA, (N)FA, R(E), (R)G or (q)uit ')
if structure_type == 'q':
return
if structure_type not in structure_letters:
print('Invalid structure selected... ')
else:
break
# ToDo
# check for consistency issues (e.g. function contains a state not in
# states
#
# discover how to convert the transition function to required format
if structure_type == 'D':
print('Creating Deterministic Finite Automaton')
states = input('Insert the states: ').split('')
input_symbols = input('Insert the input symbols: ').split('')
transition_function = input('Insert the transition function: ')
initial_state = input('Insert the initial state: ')
accept_states = input('Insert the accept states: ').split('')
d = dfa.DFA(states, input_symbols, transition_function,
initial_state, accept_states)
print(d)
loaded_structures.append(d)
return d
elif structure_type == 'E':
print('Creating Regular Expression')
expression = input('Insert the base expression: ')
r = regex.RE(expression)
print(r)
loaded_structures.append(r)
return r
elif structure_type == 'G':
print('Creating Regular Grammar')
nonterminals = input('Insert the nonterminal symbols: ').split('')
terminals = input('Insert the terminal symbols: ').split('')
productions = input('Insert the grammar productions: ')
start_symbol = input('Insert the start symbol: ')
g = rg.RG(nonterminals, terminals, productions, start_symbol)
print(g)
loaded_structures.append(g)
return g
elif structure_type == 'N':
print('Creating Nondeterministic Finite Automaton')
states = input('Insert the states: ').split('')
input_symbols = input('Insert the input symbols: ').split('')
transition_function = input('Insert the transition function: ')
initial_state = input('Insert the initial state: ')
accept_states = input('Insert the accept states: ').split('')
n = nfa.NFA(states, input_symbols, transition_function,
initial_state, accept_states)
print(n)
loaded_structures.append(n)
return n
def cnn_model_fn(features, labels, mode):
"""Model function for CNN."""
# Input Layer
# MNIST images are 28x28 pixels, and have one color channel
input_layer = input_layer = tf.reshape(features["x"], [-1, 784])
# First dense layer
# 200 nodes
with tf.variable_scope('dense1') as scope:
dim = input_layer.get_shape()[1].value
weights = tf.get_variable(name='weights',
shape=[dim, 200],
dtype=tf.float32,
initializer=tf.truncated_normal_initializer(
mean=0.0, stddev=0.2, dtype=tf.float32))
biases = tf.get_variable(name='biases',
shape=[200],
dtype=tf.float32,
initializer=tf.constant_initializer(
0.0, dtype=tf.float32))
pre_activation = tf.add(tf.matmul(input_layer, weights),
biases,
name='pre_activation')
dense1 = tf.nn.tanh(pre_activation, name='activations')
# Second dense layer
# 100 nodes
with tf.variable_scope('dense2') as scope:
dim = dense1.get_shape()[1].value
weights = tf.get_variable(name='weights',
shape=[dim, 100],
dtype=tf.float32,
initializer=tf.truncated_normal_initializer(
mean=0.0, stddev=0.2, dtype=tf.float32))
biases = tf.get_variable(name='biases',
shape=[100],
dtype=tf.float32,
initializer=tf.constant_initializer(
0.0, dtype=tf.float32))
pre_activation = tf.add(tf.matmul(dense1, weights),
biases,
name='pre_activation')
dense2 = tf.nn.tanh(pre_activation, name='activations')
# Logits layer
with tf.variable_scope('logits') as scope:
dim = dense2.get_shape()[1].value
weights = tf.get_variable(name='weights',
shape=[dim, 10],
dtype=tf.float32,
initializer=tf.truncated_normal_initializer(
mean=0.0, stddev=0.2, dtype=tf.float32))
biases = tf.get_variable(name='biases',
shape=[10],
dtype=tf.float32,
initializer=tf.constant_initializer(
0.0, dtype=tf.float32))
pre_activation = tf.add(tf.matmul(dense2, weights),
biases,
name='pre_activation')
logits = tf.add(pre_activation, 0.0, name='activations')
predictions = {
# Generate predictions (for PREDICT and EVAL mode)
"classes": tf.argmax(input=logits, axis=1),
# Add `softmax_tensor` to the graph. It is used for PREDICT and by the
# `logging_hook`.
"probabilities": tf.nn.softmax(logits, name="softmax_tensor")
}
if mode == tf.estimator.ModeKeys.PREDICT:
return tf.estimator.EstimatorSpec(mode=mode, predictions=predictions)
# Calculate Loss (for both TRAIN and EVAL modes)
loss = tf.losses.sparse_softmax_cross_entropy(labels=labels, logits=logits)
# Configure the Training Op (for TRAIN mode)
if mode == tf.estimator.ModeKeys.TRAIN:
optimizer = DFA.DFA(learning_rate=0.5)
train_op = optimizer.minimize(loss=loss,
global_step=tf.train.get_global_step())
return tf.estimator.EstimatorSpec(mode=mode,
loss=loss,
train_op=train_op)
# Add evaluation metrics (for EVAL mode)
eval_metric_ops = {
"accuracy":
tf.metrics.accuracy(labels=labels, predictions=predictions["classes"])
}
return tf.estimator.EstimatorSpec(mode=mode,
loss=loss,
eval_metric_ops=eval_metric_ops)
def get(self):
dfa_ = dfa.DFA()
return jsonify({'all_words': dfa_.all_words}), 200
3: {
'a': 3,
'b': 5
},
4: {
'a': 6,
'b': 2
},
5: {
'a': 3,
'b': 0
},
6: {
'a': 3,
'b': 1
},
}
S = 0
T = {4, 5, 6}
new_dfa = dfa.DFA(VN, VT, f, S, T)
new_dfa.minimize()
print(new_dfa.VN)
print(new_dfa.VT)
for k, v in new_dfa.f.items():
print(k, v)
print(new_dfa.S)
print(new_dfa.T)
def convert_to_dfa(self) :
dfa_state_size = 1
final_states = []
dfa_final_states = []
dfa_initial_state = 'q0'
delta_prime_func = dict()
# dfa_states just stores the name of the states. for example : ['q0', 'q1', ...]
dfa_states = [dfa_initial_state]
# since in the new dfa machine the states might be made up of several states of the nfa machine, and it might be a little bit
# hard to store the new states as a list, i used dictionary to store the states in this format
# for example : {'q0' : ['Q1', 'Q2'], 'q1' : ['Q0'], ...}
# in this example q's are dfa's states and the Q's are nfa's state, but the list of the nfa states is the equivalent
dfa_states_name = dict()
dfa_delta_func = dict()
dfa_moves = []
dfa_states_name[dfa_initial_state] = [self.initial_state]
dfa_delta_func[dfa_initial_state] = dict()
# first of all i find the new machine's final states which depend's on the nfa's final states
for state in self.states :
lamda_closure = self.lambda_closure(state)
for s in lamda_closure :
if s in self.final_states :
final_states.append(state)
# and then we use lambda_closure function to produce the delta prime transition function of the nfa state
# which helps us to produce the result dfa machine
for state in self.states :
lambda_closure = self.lambda_closure(state)
delta_prime_func[state] = dict()
for alpha in self.alphabet :
res = set()
for s in lambda_closure :
# since in nfa machine we might have no transition for a state for the given input
# i used try in order not to result in exception
try :
tmp = self.delta_func[s][alpha]
for i in tmp :
res = res.union(self.lambda_closure(i))
except :
pass
delta_prime_func[state][alpha] = res
# computing the new machine's delta transition function until there is no new state
for dfa_state in dfa_states :
for alpha in self.alphabet :
res = set()
for state in dfa_states_name[dfa_state] :
res = res.union(delta_prime_func[state][alpha])
found = False
res = list(res)
next_state = ''
for name, state in dfa_states_name.items() :
if sorted(state) == sorted(res) :
found = True
next_state = name
# if the result state is not in the dfa_states we will consider it as a new state
if not found :
new_state_name = 'q' + str(dfa_state_size)
dfa_states.append(new_state_name)
dfa_states_name[new_state_name] = res
dfa_delta_func[new_state_name] = dict()
next_state = new_state_name
dfa_state_size += 1
dfa_delta_func[dfa_state][alpha] = next_state
# computing the final states based on the new names
for name, state in dfa_states_name.items() :
for s in state :
if s in final_states and name not in dfa_final_states:
dfa_final_states.append(name)
# producing the vertices in order to call the DFA class
for current_state in dfa_states_name :
for alphabet in self.alphabet :
dfa_moves.append([current_state, alphabet, dfa_delta_func[current_state][alphabet]])
return dfa.DFA(self.alphabet, dfa_states, dfa_initial_state, dfa_final_states, dfa_moves)
import dfa
"""
Initializing a DFA.
"""
d = dfa.DFA()
"""
Add a set of symbols to the alphabet of the DFA. You can
also add individual symbols.
"""
d.add_alphabet(1, 0)
"""
Add a set of states to the DFA. You can also add individual states.
"""
d.add_states('A', 'BEF', 'CF', 'DF', 'F', 'B', 'C', 'D')
"""
Add an accepting state. If the state is already in the DFA,
the state is marked as an accepting state.
"""
d.add_accepting_state('A')
d.add_accepting_state('BEF')
d.add_accepting_state('DF')
d.add_accepting_state('F')
d.add_accepting_state('D')
d.add_accepting_state('CF')
"""
By default, PyAutomata sets the start state to a state called 'q0'.
You can change this by calling the following method.
"""
d.start_state('A')
"""
By default, PyAutomata does not include a rejecting state in each DFA.
