def branin(n_eval, random_seed):
evaluator = Branin()
space = [
hp.randint('x' + str(i), evaluator.n_vertices[i])
for i in range(len(evaluator.n_vertices))
]
init_points = sample_init_points(evaluator.n_vertices, 1,
random_seed).long().numpy()
init_points = [{
'x' + str(j): init_points[i][j]
for j in range(len(evaluator.n_vertices))
} for i in range(init_points.shape[0])]
def evaluate(x):
return evaluator.evaluate(torch.from_numpy(np.array(x))).item()
trials = base.Trials()
fmin(evaluate,
space,
algo=tpe.suggest,
max_evals=n_eval,
points_to_evaluate=init_points,
trials=trials)
evaluations, optimum = evaluations_from_trials(trials)
return optimum
def __init__(self, random_seed_pair=(None, None)):
citywise_satisfaction, tourism_attractions_similarity, transportation_type, travel_time, cost, delay_prob = generate_travel_plan_problem(
TRAVEL_N_CITIES, random_seed_pair[0])
self.n_vertices = [TRAVEL_N_CITIES] + list(
number_of_edges(transportation_type))
self.suggested_init = torch.empty(0).long()
self.suggested_init = torch.cat([
self.suggested_init,
sample_init_points(self.n_vertices,
20 - self.suggested_init.size(0),
random_seed_pair[1]).long()
],
dim=0)
self.adjacency_mat = []
self.fourier_freq = []
self.fourier_basis = []
self.random_seed_info = 'R'.join([
str(random_seed_pair[h]).zfill(4)
if random_seed_pair[h] is not None else 'None' for h in range(2)
])
for i in range(len(self.n_vertices)):
n_v = self.n_vertices[i]
adjmat = torch.diag(torch.ones(n_v - 1), -1) + torch.diag(
torch.ones(n_v - 1), 1)
self.adjacency_mat.append(adjmat)
laplacian = torch.diag(torch.sum(adjmat, dim=0)) - adjmat
eigval, eigvec = torch.symeig(laplacian, eigenvectors=True)
self.fourier_freq.append(eigval)
self.fourier_basis.append(eigvec)
self.citywise_satisfaction = citywise_satisfaction
self.tourism_attractions_similarity = tourism_attractions_similarity
self.transportation_type = transportation_type
self.travel_time = travel_time
self.cost = cost
self.delay_prob = delay_prob
def pest_control(n_eval, random_seed):
space = [
hp.choice('x' + str(i + 1).zfill(2), range(PESTCONTROL_N_CHOICE))
for i in range(PESTCONTROL_N_STAGES)
]
init_points = sample_init_points([PESTCONTROL_N_CHOICE] *
PESTCONTROL_N_STAGES, 20,
random_seed).long().numpy()
init_points = [{
'x' + str(j + 1).zfill(2): init_points[i][j]
for j in range(PESTCONTROL_N_STAGES)
} for i in range(20)]
def evaluate(x):
return _pest_control_score(np.array(x))
trials = base.Trials()
fmin(evaluate,
space,
algo=tpe.suggest,
max_evals=n_eval,
points_to_evaluate=init_points,
trials=trials)
evaluations, optimum = evaluations_from_trials(trials)
return optimum
def pest_control(n_eval, random_seed): init_points = sample_init_points([PESTCONTROL_N_CHOICE] * PESTCONTROL_N_STAGES, 20, random_seed).long().numpy() evaluations = [] for i in range(init_points.shape[0]): evaluations.append(_pest_control_score(init_points[i])) for i in range(n_eval): evaluations.append(_pest_control_score(np.random.randint(0, PESTCONTROL_N_CHOICE, (PESTCONTROL_N_STAGES, )))) evaluations, optimum = evaluations_from_list(evaluations) return optimum
def branin(n_eval, random_seed): evaluator = Branin() init_points = sample_init_points(evaluator.n_vertices, 1, random_seed).long().numpy() def evaluate(x): return evaluator.evaluate(torch.from_numpy(x)).item() evaluations = [] for i in range(init_points.shape[0]): evaluations.append(evaluate(init_points[i])) for i in range(n_eval): evaluations.append(evaluate(np.array([np.random.randint(0, evaluator.n_vertices[elm]) for elm in range(len(evaluator.n_vertices))]))) evaluations, optimum = evaluations_from_list(evaluations) return optimum
def centroid(n_eval, random_seed_pair): evaluator = Centroid(random_seed_pair) init_points = sample_init_points([CENTROID_N_CHOICE] * CENTROID_N_EDGES, 20, random_seed_pair[1]).long().numpy() def evaluate(x): return evaluator.evaluate(torch.from_numpy(x)).item() evaluations = [] for i in range(init_points.shape[0]): evaluations.append(evaluate(init_points[i])) for i in range(n_eval): evaluations.append(evaluate(np.random.randint(0, CENTROID_N_CHOICE, (CENTROID_N_EDGES,)))) evaluations, optimum = evaluations_from_list(evaluations) return optimum
def centroid(n_eval, random_seed_pair):
space = [
hp.choice('x' + str(i + 1).zfill(2), range(CENTROID_N_CHOICE))
for i in range(CENTROID_N_EDGES)
]
init_points = sample_init_points([CENTROID_N_CHOICE] * CENTROID_N_EDGES,
20, random_seed_pair[1]).long().numpy()
init_points = [{
'x' + str(j + 1).zfill(2): init_points[i][j]
for j in range(CENTROID_N_EDGES)
} for i in range(20)]
evaluator = Centroid(random_seed_pair)
def evaluate(x):
interaction_mixed = edge_choice(np.array(x),
evaluator.interaction_list)
partition_mixed = partition(interaction_mixed, CENTROID_GRID)
kld_sum = 0
for i in range(evaluator.n_ising_models):
kld = ising_dense(
interaction_sparsified=interaction_mixed,
interaction_original=evaluator.interaction_list[i],
covariance=evaluator.covariance_list[i],
partition_sparsified=partition_mixed,
partition_original=evaluator.partition_original_list[i],
grid_h=CENTROID_GRID[0])
kld_sum += kld
return kld_sum / float(evaluator.n_ising_models)
trials = base.Trials()
fmin(evaluate,
space,
algo=tpe.suggest,
max_evals=n_eval,
points_to_evaluate=init_points,
trials=trials)
evaluations, optimum = evaluations_from_trials(trials)
return optimum