def test_preffered(self): swaps = list(sgp.swaps(self.solution)) #print(swaps) violats = list(map(lambda t: (sgp.evaluate_swap(self.solution, t), t), swaps)) print(min(violats), max(violats)) """self.assertEqual(
def learning(self, solution):
tabu = [ {} for week in solution ]
best_solution = copy.deepcopy(solution)
best_violations = sgp.violations(solution)[0]
# wie lange wird weiter gesucht, obwohl keine verbesserung eintritt
#learning_semaphore = self.max_learning
for iteration in range(self.groups * self.weeks):
if best_violations == 0 or self.timeout: # were done
break
try:
# von allen Täuschen, den besten finden
nex_viol, w, a, b = min(
filter(
# entweder verbessert dieser Tausch diese Lösung zur Besten
# oder er ist nicht tabu
lambda t: t[0] < best_violations or not tuple(sorted((t[2], t[3]))) in tabu[t[1]] or tabu[t[1]][tuple(sorted((t[2], t[3])))] < iteration,
map(
lambda t: (sgp.evaluate_swap(solution, t), t[0], t[1], t[2]),
sgp.swaps(solution)
)
)
)
sgp.swap(solution, w, a, b)
tabu[w][tuple(sorted((a, b)))] = iteration + random.randint(4, 100)
if nex_viol < best_violations:
best_solution = solution
best_violations = nex_viol
print(best_violations, end=' ')
sys.stdout.flush()
except ValueError as e:
break
return best_solution