def test_random_optimise(self):
""" Test direct optmisation."""
self.report('Rand optimise:')
random_minimise = lambda obj, bounds, max_evals: oper_utils.random_minimise(
obj, bounds, max_evals, False, False)
random_maximise = lambda obj, bounds, max_evals: oper_utils.random_maximise(
obj, bounds, max_evals, False, False)
self._test_optimise_method(random_minimise, random_maximise, False)
def _rand_maximise_obj(self, obj, num_evals):
""" Maximise with random evaluations. """
if num_evals is None:
lead_const = 10 * min(5, self.dim)**2
num_evals = lambda t: np.clip(lead_const * np.sqrt(min(t, 1000)),
2000, 3e4)
opt_val, opt_pt = random_maximise(obj, self.bounds, num_evals)
return opt_val, opt_pt
def test_random_maximise(self):
""" Test direct optmisation."""
self.report('Rand maximise:')
num_max_successes = 0
for prob in self.problems:
prob_bounds = np.concatenate(
(np.array(prob.lb).reshape(1, -1), np.array(prob.ub).reshape(
1, -1)),
axis=0).T
max_val_soln, _ = oper_utils.random_maximise(
prob.obj, prob_bounds, 2 * self.max_evals, False)
diff = abs(prob.max_val - max_val_soln)
self.report(
prob.descr + '(max):: True: %0.4f, Soln: %0.4f, diff: %0.4f' %
(prob.max_val, max_val_soln, diff), 'test_result')
max_is_successful = diff < 1e-3
num_max_successes += max_is_successful
# Check if successful
assert num_max_successes >= 1
def _rand_wrap(*args): """ A wrapper so as to only return the optimal point. """ _, opt_pt = random_maximise(*args, vectorised=False) return opt_pt