def pick_one(self):
index = 0
r = random()
while r > 0:
r -= self.paths[index].prob
index += 1
new_path = Path(self.paths[index - 1].nodes)
new_path.normalize_score = self.paths[index - 1].normalize_score
new_path.prob = self.paths[index - 1].prob
return new_path
def testConvertDHCtoUHC():
from uhc import uhc
from dhc import dhc
from graph import Path
instances = [
'',
'a,a',
'a,b',
'a,b b,a',
'a,b b,c c,a',
'a,b b,c a,c',
'a,b b,c c,d',
'a,b b,c c,d d,a',
'a,b b,c c,d a,d',
'a,b b,c c,d a,d d,b c,a',
]
for instance in instances:
convertedInstance = convertDHCtoUHC(instance)
instanceSolution = dhc(instance)
convertedInstanceSolution = uhc(convertedInstance)
revertedSolution = revertSolution(convertedInstanceSolution)
utils.tprint(instance, 'maps to', convertedInstance,\
' solutions were: ', instanceSolution, ';', convertedInstanceSolution)
utils.tprint('revertedSolution', revertedSolution)
if revertedSolution == 'no':
assert instanceSolution == 'no'
else:
g = Graph(instance, weighted=False)
path = Path.fromString(revertedSolution)
# print('g', g, 'path', path)
assert g.isHamiltonCycle(path) or g.isHamiltonCycle(path.reverse())
def testTspDirK():
testvals = [
(';1', 'no'),
('a,a,3;1', 3),
('a,a,3 ; 1 ', 3),
('a,a,3;2', 'no'),
('a,b,4;1', 'no'),
('a,b,4;2', 'no'),
('a,b,4;3', 'no'),
('a,b,4 b,a,9;1', 13),
('a,b,4 b,a,9;2', 13),
('a,b,4 b,a,9;3', 'no'),
('a,b,4 b,a,9 b,c,6 c,a,10;2', 13),
('a,b,4 b,a,9 b,c,6 c,a,10;3', 20),
('a,b,4 b,a,9 b,c,6 c,a,1;2', 11),
('a,b,4 b,a,9 b,c,6 c,a,10;4', 'no'),
('a,b,4 b,a,9 b,c,6 c,a,10 a,c,2 c,b,3;3', 14),
('a,b,4 b,a,9 b,c,6 c,a,10 a,c,2 c,b,3;4', 'no'),
]
for (inString, solution) in testvals:
val = tspDirK(inString)
utils.tprint(inString.strip(), ':', val)
if solution == 'no':
assert val == solution
else:
(graphString, K) = inString.split(';')
graph = Graph(graphString, directed=True)
K = int(K)
cycle = Path.fromString(val)
dist = graph.cycleLength(cycle)
assert graph.isCycle(cycle)
assert len(cycle) >= K
assert dist == solution
def choose_path_for_ant(self, pheromone_influence, heuristic_influence):
chosen_edges = []
node = self.graph.start
while not (node == self.graph.end):
possible_edges = self.graph.edges_adjacent_to_node[node][:]
if (chosen_edges): possible_edges.remove(chosen_edges[-1])
# Calculate the probabilities.
# If another ant in current iteration has already been in this node, having come
# through the same edge, then the probabilities are already calculated.
if (chosen_edges
and node in self.edge_choice_probabilities
and chosen_edges[-1] in self.edge_choice_probabilities[node]
):
probabilities = self.edge_choice_probabilities[node][chosen_edges[-1]]
if (self.verbosity == 2):
for i in range (len(possible_edges)):
print (possible_edges[i], "[prob:", "{:.3f}".format(probabilities[i]), end="], ")
else:
sum = 0
for edge in possible_edges:
sum += self.pheromone_array[edge]**pheromone_influence
sum += (1/self.graph.edges_lenght[edge])**heuristic_influence
probabilities = []
for edge in possible_edges:
probabilities.append((self.pheromone_array[edge]**pheromone_influence \
+ (1/self.graph.edges_lenght[edge])**heuristic_influence) \
/ sum)
if (self.verbosity == 2):
print (edge, "[prob:", "{:.3f}".format(probabilities[-1]), end="], ")
# Store the result for future ants.
if (chosen_edges):
if (not node in self.edge_choice_probabilities):
self.edge_choice_probabilities[node] = {}
self.edge_choice_probabilities[node][chosen_edges[-1]] = probabilities
# Draw the next edge.
drawn_edge = random.choices(possible_edges, probabilities, k=1)[0]
if (self.verbosity == 2):
print("drawn edge: ", drawn_edge)
chosen_edges.append(drawn_edge)
if (node == drawn_edge[0]):
node = drawn_edge[1]
else:
assert (node == drawn_edge[1])
node = drawn_edge[0]
assert (chosen_edges)
lenght = 0
for edge in chosen_edges: lenght += self.graph.edges_lenght[edge]
if (self.verbosity >= 1):
if (self.verbosity == 2): print()
print ("Chosen path: ", chosen_edges, "\nLenght: ", lenght, end="\n\n")
return Path(chosen_edges, lenght)
def tspDir(inString):
graph = Graph(inString)
firstNode = graph.chooseNode()
if firstNode == None: # graph contains no vertices
return '' # this is a positive instance -- we return the empty cycle
(cycle, distance) = shortestCycleWithPrefix(graph, Path([firstNode]), 0)
if cycle != None:
return str(cycle)
else:
return 'no'
def find_best_route(graph, start, end):
"""
Find the best route through the grid between the given start node and the given
end node.
"""
if start == end:
best_route_found = True
else:
best_route_found = False
graph.nodes[start].distance = 0
reachable_unprocessed_nodes = set([graph.nodes[start]])
while not best_route_found:
# 1. loop over the reachable but unprocessed nodes. Find the one with the
# least distance from the start (X).
# 2. For each node Y connected to X, take d(start,X) + d(X,Y) and see if it
# is less than d(start, Y). If so, update node Y with a new distance and
# a new previous node (X).
# 3. If the end has been reached, iterate over all items in the reachable
# but unprocessed nodes. If d(start,end) <= d(start,node) for all of these
# nodes, then we have found the best route.
# Find the unprocessed node closest to the start
closest_distance = None
for node in reachable_unprocessed_nodes:
if closest_distance is None or node.distance < closest_distance:
closest_node = node
closest_distance = node.distance
assert closest_node.processed == False, \
"Processed node is in list of unprocessed nodes"
if closest_node.node_id == end:
best_route_found = True
# Process this node
if not best_route_found:
reachable_unprocessed_nodes.remove(closest_node)
closest_node.processed = True
for (node_id, distance) in closest_node.connections:
node = graph.nodes[node_id]
if node.distance is None or closest_node.distance + distance < node.distance:
node.distance = closest_node.distance + distance
reachable_unprocessed_nodes.add(node)
node.previous_node = closest_node
# Create the path of the best route.
current_node = graph.nodes[end]
best_route = [current_node]
if start != end:
while current_node.previous_node is not None:
current_node = current_node.previous_node
best_route.insert(0, current_node)
return Path(best_route)
def tspPath(inString):
graphStr, source, dest = [x.strip() for x in inString.split(";")]
graph = Graph(graphStr, weighted=True, directed=False)
# If there are two vertices, our conversion won't work correctly,
# so treat this as a special case.
if len(graph) == 2:
if graph.containsEdge(Edge([source, dest])):
return str(Path([source, dest]))
else:
return 'no'
# add an overwhelmingly negative edge from source to dest, thus
# forcing that to be part of a shortest Ham cycle.
sumOfWeights = graph.sumEdgeWeights()
fakeEdge = Edge([source, dest])
if graph.containsEdge(fakeEdge) == True:
graph.removeEdge(fakeEdge)
graph.addEdge(fakeEdge, -sumOfWeights)
tspSoln = tsp(str(graph))
# print('graph', graph)
# print('tspSoln', tspSoln)
if tspSoln == 'no':
return 'no'
# convert from string to Path object
tspSoln = Path.fromString(tspSoln)
rotatedSoln = tspSoln.rotateToFront(source)
if rotatedSoln[-1] != dest:
# Probably oriented the wrong way (or else fakeEdge wasn't
# used -- see below). Reverse then rotate source to front
# again
rotatedSoln = rotatedSoln.reverse().rotateToFront(source)
# If dest still isn't at the end of the cycle, then the orginal graph didn't
# have a Ham path from source to dest (but it did have a Ham
# cycle -- a cycle that did *not* include fakeEdge).
if rotatedSoln[-1] != dest:
return 'no'
else:
return str(rotatedSoln)
def testShortestCycleWithPrefix():
graph = Graph('''
a,b,5 b,a,3 b,c,6
c,a,7 c,d,8 d,a,9 a,c,1 d,b,2
''',
directed=True)
path = Path(['a'])
val = shortestCycleWithPrefix(graph, path, 0)
cycle, dist = val
utils.tprint(val)
assert graph.isHamiltonCycle(cycle)
assert dist == 14
def bruteForceSearch(digraph, start, end, maxTotalDist, maxDistOutdoors):
"""
Finds the shortest path from start to end using brute-force approach.
The total distance travelled on the path must not exceed maxTotalDist, and
the distance spent outdoor on this path must not exceed maxDisOutdoors.
Parameters:
digraph: instance of class Digraph or its subclass
start, end: start & end building numbers (strings)
maxTotalDist : maximum total distance on a path (integer)
maxDistOutdoors: maximum distance spent outdoors on a path (integer)
Assumes:
start and end are numbers for existing buildings in graph
Returns:
The shortest-path from start to end, represented by
a list of building numbers (in strings), [n_1, n_2, ..., n_k],
where there exists an edge from n_i to n_(i+1) in digraph,
for all 1 <= i < k.
If there exists no path that satisfies maxTotalDist and
maxDistOutdoors constraints, then raises a ValueError.
"""
stack = [Path([SmartNode(start)], 0, 0)]
valid_paths = []
while stack:
path = stack.pop(-1)
for edge in digraph.childrenOf(path.get_current_position()):
if not path.is_node_visited(edge.getDestination()):
new_path = path.clone()
new_path.add_edge(edge)
# "...consider first finding all valid paths that satisfy
# the max distance outdoors constraint,..."
if new_path.distance_outdoors <= maxDistOutdoors:
if new_path.is_end_reached(end):
valid_paths.append(new_path)
else:
stack.append(new_path)
# "... and then going through those paths and returning the shortest,
# rather than trying to fulfill both constraints at once..."
valid_paths = filter(lambda path: path.total_distance <= maxTotalDist,
valid_paths)
# min will raise a ValueError if an empty sequence is passed to it
shortest = min(valid_paths, key=lambda x: x.total_distance)
return shortest.get_node_list()
def verifyTspD(I, S, H):
if S == 'no': return 'unsure'
# extract G,L from I, and convert to correct data types etc.
(G, L) = I.split(';')
G = Graph(G, directed=False)
L = int(L)
# split the hint string into a list of vertices, which will
# form a Hamilton cycle of length at most L, if the hint is correct
cycle = Path.fromString(H)
# verify the hint is a Hamilton cycle, and has length at most L
if G.isHamiltonCycle(cycle) and \
G.cycleLength(cycle) <= L:
return 'correct'
else:
return 'unsure'
def _pathN2G(self, neopath):
path = Path()
n0 = self._nodeN2G(neopath.start_node)
for rel in neopath:
relation = self._relN2G(rel)
path.relations.append(relation)
for n in neopath.nodes:
path.nodes.append(self._nodeN2G(n))
for i in range(len(path.nodes)):
path.elements.append(path.nodes[i])
if path.relations and i < len(path.relations):
path.elements.append(path.relations[i])
return path
def testTsp():
testvals = [
('a,b,5', 'no', None),
('a,b,5 b,c,6 c,d,8 d,a,9 a,c,1 d,b,2', 'a,b,d,c', 16),
('a,b,5 b,c,6 c,d,8 d,a,9 a,c,1 d,b,200', 'a,b,c,d', 28),
('a,b,5 b,c,6 d,a,9 a,c,1', 'no', None),
]
for (inString, solution, length) in testvals:
val = tsp(inString)
utils.tprint(inString, ':', val)
if solution == 'no':
assert val == solution
else:
g = Graph(inString, weighted=True, directed=False)
p = Path.fromString(val)
assert g.isHamiltonCycle(p)
assert g.cycleLength(p) == length
def tspDirK(inString):
(graphString, K) = inString.split(';')
K = int(K)
# treat K=0 as a special case. Since the empty cycle satisfies
# this, we return a positive solution which is the empty string.
if K==0:
return ''
graph = Graph(graphString)
bestCycle = None; bestDistance = None
for firstNode in graph:
(cycle, distance) = shortestKCycleWithPrefix(graph, K, Path([firstNode]), 0)
if cycle != None:
if bestDistance == None or distance < bestDistance:
(bestCycle, bestDistance) = (cycle, distance)
if bestCycle != None:
return str(bestCycle)
else:
return 'no'
def testUhc():
testVals = [
('', True),
('a,b', False),
('a,b b,c', False),
('a,b b,c c,a', True),
('a,b b,c c,a a,d', False),
('a,b b,c c,a a,d b,d', True),
('aB,aA aB,aC aA,aC', True),
]
for (inString, hasUhc) in testVals:
result = uhc(inString)
utils.tprint(inString, ':', result)
if not hasUhc:
assert result == 'no'
else:
g = Graph(inString, weighted=False, directed=False)
path = Path.fromString(result)
assert g.isHamiltonCycle(path)
def directedDFS(digraph, start, end, maxTotalDist, maxDistOutdoors):
"""
Finds the shortest path from start to end using directed depth-first.
search approach. The total distance travelled on the path must not
exceed maxTotalDist, and the distance spent outdoor on this path must
not exceed maxDisOutdoors.
Parameters:
digraph: instance of class Digraph or its subclass
start, end: start & end building numbers (strings)
maxTotalDist : maximum total distance on a path (integer)
maxDistOutdoors: maximum distance spent outdoors on a path (integer)
Assumes:
start and end are numbers for existing buildings in graph
Returns:
The shortest-path from start to end, represented by
a list of building numbers (in strings), [n_1, n_2, ..., n_k],
where there exists an edge from n_i to n_(i+1) in digraph,
for all 1 <= i < k.
If there exists no path that satisfies maxTotalDist and
maxDistOutdoors constraints, then raises a ValueError.
"""
stack = [Path([SmartNode(start)], 0, 0)]
while stack:
path = stack.pop(-1)
for edge in digraph.childrenOf(path.get_current_position()):
if not path.is_node_visited(edge.getDestination()):
new_path = path.clone()
new_path.add_edge(edge)
if (new_path.distance_outdoors <= maxDistOutdoors
and new_path.total_distance <= maxTotalDist):
if new_path.is_end_reached(end):
return new_path.get_node_list()
else:
stack.append(new_path)
raise ValueError()
def testTspPath():
testvals = [
('a,b,5;a;b', 5),
('a,b,5 b,c,6 c,d,8 d,a,9 a,c,1 d,b,2;a;b', 11),
('a,b,5 b,c,6 c,d,8 d,a,9 a,c,1 d,b,2 c,e,10 d,e,20;a;b', 33),
('a,b,5 b,c,6 d,a,9 a,c,1 d,b,2;a;b', 'no'),
('a,b,5 b,c,6 d,a,9 a,c,1;a;b', 'no'),
]
for (inString, solution) in testvals:
val = tspPath(inString)
utils.tprint(inString.strip(), ':', val)
if solution == 'no':
assert val == solution
else:
graphStr, source, dest = [x.strip() for x in inString.split(";")]
graph = Graph(graphStr, directed=False)
path = Path.fromString(val)
dist = graph.pathLength(path)
assert graph.isHamiltonPath(path)
assert dist == solution
def perform_algorithm(self, iterations, population_size, evaporation_factor,
pheromone_influence, heuristic_influence):
self.initialize_pheromone_array()
self.edge_choice_probabilities = {}
shortest_path = Path([], float("inf"))
for i in range(iterations):
if (self.verbosity >= 1):
print ("=== Iteration ", i, end="\n\n")
self.edge_choice_probabilities.clear()
for j in range(population_size):
if (self.verbosity >= 1):
print ("Ant ", j)
path = self.choose_path_for_ant(pheromone_influence, heuristic_influence)
self.drop_pheromone_on_path(path)
if (path.lenght < shortest_path.lenght):
shortest_path = path
self.update_pheromone_array(evaporation_factor, shortest_path)
if (i % self.state_displaying_period == 0):
self.printer.display_state(self.pheromone_array, i)
self.printer.display_state(self.pheromone_array, iterations)
def verifyTspDPolytime(I, S, H):
# reject excessively long solutions and hints
if len(S) > len(I) or len(H) > len(I):
return 'unsure'
# The remainder of the program is identical to verifyTspD.py
# ...
if S == 'no': return 'unsure'
# extract G,L from I, and convert to correct data types etc.
(G, L) = I.split(';')
G = Graph(G, directed=False)
L = int(L)
# split the hint string into a list of vertices, which will
# form a Hamilton cycle of length at most L, if the hint is correct
cycle = Path.fromString(H)
# verify the hint is a Hamilton cycle, and has length at most L
if G.isHamiltonCycle(cycle) and \
G.cycleLength(cycle) <= L:
return 'correct'
else:
return 'unsure'
def testTspDir():
testvals = [
('aC,aB,1 aA,aB,1 aC,aA,1 aB,aA,1 aB,aC,1 aA,aC,1', 3),
('', 0),
('a,a,3', 3),
('a,b,4', 'no'),
('a,b,4 b,a,9', 13),
('a,b,4 b,a,9 b,c,6 c,a,10', 20),
('a,b,4 b,a,9 b,c,6 c,a,10 a,c,2 c,b,3', 14),
('a,b,4 b,a,9 b,c,6 c,a,10 a,c,2 c,b,300', 20),
('a,b,4 b,a,9 b,c,6 c,a,10 a,c,2 c,b,3 c,d,1', 'no'),
]
for (inString, solution) in testvals:
val = tspDir(inString)
utils.tprint(inString.strip(), ':', val)
if solution == 'no':
assert val == solution
else:
graph = Graph(inString, directed=True)
cycle = Path.fromString(val)
dist = graph.cycleLength(cycle)
assert graph.isHamiltonCycle(cycle)
assert dist == solution
def testHalfUhc():
testvals = [
('', 'yes'),
('a,b', 'no'),
('a,a', 'yes'),
('a,b b,c', 'no'),
('a,b b,c c,a', 'yes'),
('a,b b,c c,a d,e d,f', 'yes'),
('a,b b,c c,a d,e f,g', 'no'),
('a,b b,c c,a a,d', 'yes'),
('a,b b,c c,a a,d b,d', 'yes'),
('a,b b,c c,d d,e e,f f,g g,h h,i i,j h,a', 'yes'),
('a,b b,c c,d d,e e,f f,g g,h h,i i,j d,a', 'no'),
]
for (inString, solution) in testvals:
val = halfUhc(inString)
utils.tprint(inString, ':', val)
if solution == 'no':
assert val == solution
else:
g = Graph(inString, weighted=False, directed=False)
path = Path.fromString(val)
assert g.isCycle(path)
assert g.cycleLength(path) >= len(g) / 2
def generate_initial_population(self):
for i in range(self.popsize):
new_path = Path(self.graph.get_random_path(self.start, self.end))
self.paths.append(new_path)
