class GA_TSP:
#object to hold graph
graph = TSP_Graph("",0)
#size of the population
npop = 0
#max number of generation
max_g = 100;
#current population
pop = []
#mating pool
mating_pool = []
#store the average cost of each generation
average_cost = []
#min cost of each generation
min_cost = []
modified_pop_cost = 0
best_tour = Tour()
def __init__(self, filename):
self.graph = TSP_Graph(filename, 15)
self.graph._read_data()
self.npop = self.graph.vcount * 2
def _create_init_pop(self):
var = []
for i in range(0,self.npop):
temp_tour = Tour()
temp = np.random.permutation(self.graph.vcount)
temp_tour.tour = temp
temp_tour.tour_cost = self.graph._get_tour_cost(temp_tour.tour)
var.append(temp_tour)
self.pop =var;
self._modify_costs()
self._sort_tours(self.pop)
def _sort_tours(self, pop):
self.pop = sorted(pop, key=lambda x: x.tour_cost)
def _modify_costs(self):
sum_cost = 0;
min_cost = sys.maxint
self.modified_pop_cost = 0
for i in range(0,self.npop):
sum_cost = sum_cost + self.pop[i].tour_cost
if(min_cost > self.pop[i].tour_cost):
min_cost = self.pop[i].tour_cost
self.best_tour = self.pop[i]
for i in range(0,self.npop):
self.pop[i].tour_cost = sum_cost - self.pop[i].tour_cost
self.modified_pop_cost = self.modified_pop_cost + self.pop[i].tour_cost
self.average_cost.append(sum_cost/self.npop)
self.min_cost.append(min_cost)
def _create_mating_pool(self):
for i in range(self.npop):
rand = np.random.randint(0, self.modified_pop_cost + 1)
for t in self.pop :
rand = rand - t.tour_cost
if(rand <= 0):
self.mating_pool.append(t)
break
#print("printing pop")
#self._print_pool(self.pop)
#print("print mate pop")
#self._print_pool(self.mating_pool)
def _get_next_gen(self):
self._create_mating_pool()
new_pop = []
temp = self.npop/2
for i in range(0,temp):
a,b = np.random.randint(0, self.npop, 2)
#print(i,a,b)
child1 = self._get_child(self.mating_pool[a], self.mating_pool[b] )
child2 = self._get_child(self.mating_pool[b], self.mating_pool[a] )
new_pop.append(child1)
new_pop.append(child2)
self.pop = new_pop
self._modify_costs()
self._sort_tours(self.pop)
self.mating_pool = []
def _get_child(self, p1,p2):
child = self._mate_parents(p1,p2)
return child
def _mate_parents(self,p1,p2):
size = self.graph.vcount
child_tour = np.zeros(self.graph.vcount) + p2.tour
while(1) :
a = np.random.randint(0, size, 2)
if(a[0]!=a[1]):
temp = min(a[0],a[1])
a[1] = a[0]+a[1] - temp
a[0] = temp
break;
index_list = []
for i in range(a[0],a[1] + 1):
temp = nonzero(p2.tour == p1.tour[i] )
index_list.append(temp[0][0])
index_list.sort()
for i in range(index_list.__len__()):
temp = a[0] + i
child_tour[index_list[i]] = p1.tour[temp]
child = Tour()
child.tour = child_tour
child.tour_cost = self.graph._get_tour_cost(child_tour)
return child
def _run(self):
self._create_init_pop();
for i in range(0,self.max_g):
#print(self.npop,self.pop.__len__(),self.mating_pool.__len__())
self._get_next_gen()
print("generation " , i )
return self.best_tour
def _print_pool(self, pool):
for t in pool:
print(t.tour.tolist(), t.tour_cost)
def __init__(self, filename):
self.graph = TSP_Graph(filename, 15)
self.graph._read_data()
self.npop = self.graph.vcount * 2
