def aStar(self):
''' START FUNCTION POINTERS '''
hashed = self.model.hashedarray
get_score = self.get_cost
set_score = self.set_cost
heuri = self.heuristic
get_total = self.get_F
network = self.find_greedy_network
is_goal = self.is_goal
best_score = lambda x, p: -self.model.bic_score(
x, self.find_best_parents(x, p))
write = self.out.write
''' END FUNCTION POINTERS '''
init_time = clock()
self.init_astar()
visited = {}
ancestor = {}
q = pqueue()
start = []
set_score(start, 0)
ancestor[hashed(start)] = (None, None)
q.put_nowait((get_total(start), start))
expanded = 0
generated = 0
while not q.empty():
F, U = q.get_nowait()
#print U , F
if is_goal(U):
cpu_time = clock() - init_time
U = self.get_path(ancestor)
write("ORDER = %s\n" % ','.join(U))
write("SCORE = %s\n" % get_score(U))
write("TIME = %s\n" % cpu_time)
write("EXPANDED = %s\n" % expanded)
write("GENERATED = %s\n" % generated)
return network(U)
if compare(get_total(U), F) < 0: continue
expanded += 1
visited[hashed(U)] = True
for X in self.data.fields:
if X in U: continue
new_U = copy(U)
new_U.append(X)
hu = hashed(new_U)
g = get_score(U) + best_score(X, U)
if hu in visited: continue
if hu not in self.g or compare(g, get_score(new_U)) < 0:
generated += 1
q.put_nowait((g + heuri(new_U), new_U))
set_score(new_U, g)
ancestor[hu] = (hashed(U), X)
return None
def aStar( self ) :
''' START FUNCTION POINTERS '''
hashed = self.model.hashedarray
get_score = self.get_cost
set_score = self.set_cost
heuri = self.heuristic
get_total = self.get_F
network = self.find_greedy_network
is_goal = self.is_goal
best_score = lambda x , p : -self.model.bic_score( x , self.find_best_parents( x , p ) )
write = self.out.write
''' END FUNCTION POINTERS '''
init_time = clock()
self.init_astar()
visited = {}
ancestor = {}
q = pqueue()
start = []
set_score( start , 0 )
ancestor[ hashed( start ) ] = ( None , None )
q.put_nowait( ( get_total( start ) , start ) )
expanded = 0
generated = 0
while not q.empty() :
F , U = q.get_nowait()
#print U , F
if is_goal( U ) :
cpu_time = clock() - init_time
U = self.get_path( ancestor )
write( "ORDER = %s\n" % ','.join( U ) )
write( "SCORE = %s\n" % get_score( U ) )
write( "TIME = %s\n" % cpu_time )
write( "EXPANDED = %s\n" % expanded )
write( "GENERATED = %s\n" % generated )
return network( U )
if compare( get_total( U ) , F ) < 0 : continue
expanded += 1
visited[ hashed( U ) ] = True
for X in self.data.fields :
if X in U : continue
new_U = copy( U )
new_U.append( X )
hu = hashed( new_U )
g = get_score( U ) + best_score( X , U )
if hu in visited : continue
if hu not in self.g or compare( g , get_score( new_U ) ) < 0 :
generated += 1
q.put_nowait( ( g + heuri( new_U ) , new_U ) )
set_score( new_U , g )
ancestor[ hu ] = ( hashed( U ) , X )
return None
def astar(m, start, end, debug):
filter_mat = sum_filter(ROBO_DIAMETER, ROBO_DIAMETER)
#valid_map = scipy.ndimage.filters.gaussian_filter(valid_pixels(m, VALID_THRESHOLD), 5.0)
valid_map = scipy.signal.convolve2d(valid_pixels(m, VALID_THRESHOLD), filter_mat, 'same')
if debug:
imgplot = plt.imshow(valid_map)
plt.show()
q = pqueue()
#put into the queue a weight of the distance plus the heuristics
#and a tuple of the current distance, and the next node to expand
if not is_valid(start, valid_map):
print "Finding new start"
start = nearest_good_cell(valid_map, start)
assert(start != None)
print "Found new start"
q.put((0, (0, start)))
height = len(m)
width = len(m[0])
closed = set()
closed.add(start)
parents = {}
path_found = False
while not q.empty():
(weight, (dist, cell)) = q.get()
if cell == end:
path_found = True
break
for neighbor in valid_neighbors(cell, width, height):
if is_valid(neighbor, valid_map) and (not neighbor in closed):
weight = dist + 1 + heuristic(neighbor, end)
new_dist = dist + 1
q.put((weight, (new_dist, neighbor)))
closed.add(neighbor)
parents[neighbor] = cell
if path_found:
cur_cell = end
path = [end]
while cur_cell != start:
cur_cell = parents[cur_cell]
path.append(cur_cell)
path.reverse()
return path
else:
print >> stderr, "NO PATH"
return []
def UCS(sentence, model):
frontier = pqueue()
for root in sentence[0]:
node = [root]
cost = model.score(node)
frontier.put((cost, node))
while (True):
if frontier.empty():
return None
else:
node = list(frontier.get()[1])
if len(node) == len(sentence):
return node, cost
else:
index = len(node)
for nextWord in sentence[index]:
newNode = node + [nextWord]
cost = model.score(newNode)
frontier.put((cost, newNode))
def UCS(sentence, model): frontier = pqueue() for root in sentence[0]: node = [root] cost = model.score(node) frontier.put((cost, node)) while(True): if frontier.empty(): return None else: node = list(frontier.get()[1]) if len(node) == len(sentence): return node, cost else: index = len(node) for nextWord in sentence[index]: newNode = node + [nextWord] cost = model.score(newNode) frontier.put((cost, newNode))
def __init__( self ):
super( PriorityQueue, self ).__init__()
self.messages = pqueue()
