def __init__(self, order, debug=False):
self.order = int(order)
self.time = 0
self.states = DirectedWeightedGraph()
self.state = ()
self.learn_state = ()
def __init__(self, order, debug=False):
self.order = int(order)
self.time = 0
self.states = DirectedWeightedGraph()
self.state = ()
self.learn_state = ()
class MarkovChain:
"""A Markov source of order n.
"""
def __init__(self, order, debug=False):
self.order = int(order)
self.time = 0
self.states = DirectedWeightedGraph()
self.state = ()
self.learn_state = ()
def set_random_state(self):
self.state = random.choice(list(self.states.V))
def next_state(self):
self.time += 1
cands = list(self.states.get_edges_at(self.state))
if len(cands) > 0:
edge = weighted_choice(cands, [c[1] for c in cands])
self.state, _ = edge
else:
self.state = None
def emit(self):
while True:
self.next_state()
if self.state is None:
raise StopIteration
yield self.state[-1]
def add_transition(self, src, dst):
"""Add a state transition into the state graph.
This will add edges with weight 1 or increase the weight if the edge
already exists.
"""
self.states.add_vertex(src)
self.states.add_vertex(dst)
self.states.add_edge(src, dst, 1)
def learn(self, source):
"""Build a markov model from an iterable input source.
"""
state = collections.deque(self.learn_state)
for i in source:
oldstate = tuple(state)
if len(oldstate) == self.order:
state.popleft()
state.append(i)
# FIXME how to handle the start case?
self.add_transition(oldstate, tuple(state))
self.learn_state = tuple(state)
def __iadd__(self, other):
if self.order != other.order:
raise ValueError(
"Cannot merge chains of different order ({} != {})".format(
self.order,
other.order
))
# FIXME: handle time and current state etc.
self.states += other.states
return self
class MarkovChain:
"""A Markov source of order n.
"""
def __init__(self, order, debug=False):
self.order = int(order)
self.time = 0
self.states = DirectedWeightedGraph()
self.state = ()
self.learn_state = ()
def set_random_state(self):
self.state = random.choice(list(self.states.V))
def next_state(self):
self.time += 1
cands = list(self.states.get_edges_at(self.state))
if len(cands) > 0:
edge = weighted_choice(cands, [c[1] for c in cands])
self.state, _ = edge
else:
self.state = None
def emit(self):
while True:
self.next_state()
if self.state is None:
raise StopIteration
yield self.state[-1]
def add_transition(self, src, dst):
"""Add a state transition into the state graph.
This will add edges with weight 1 or increase the weight if the edge
already exists.
"""
self.states.add_vertex(src)
self.states.add_vertex(dst)
self.states.add_edge(src, dst, 1)
def learn(self, source):
"""Build a markov model from an iterable input source.
"""
state = collections.deque(self.learn_state)
for i in source:
oldstate = tuple(state)
if len(oldstate) == self.order:
state.popleft()
state.append(i)
# FIXME how to handle the start case?
self.add_transition(oldstate, tuple(state))
self.learn_state = tuple(state)
def __iadd__(self, other):
if self.order != other.order:
raise ValueError("Cannot merge chains of different order ({} != {})".format(self.order, other.order))
# FIXME: handle time and current state etc.
self.states += other.states
return self
