def test_accepts_equal():
"""
Tests if the hill climbing method accepts a solution that results in the
same objective value.
"""
hill_climbing = HillClimbing()
assert_(hill_climbing.accept(rnd.RandomState(), Zero(), Zero(), Zero()))
def test_does_not_raise():
"""
This set of parameters, on the other hand, should work correctly.
"""
alns = get_alns_instance([lambda state, rnd: One()],
[lambda state, rnd: One()])
alns.iterate(Zero(), [1, 1, 1, 1], .5, HillClimbing(), 100)
# 0 and 1 are both acceptable decay parameters (since v1.2.0).
alns.iterate(Zero(), [1, 1, 1, 1], 0., HillClimbing(), 100)
alns.iterate(Zero(), [1, 1, 1, 1], 1., HillClimbing(), 100)
def test_raises_non_positive_weights():
"""
The passed-in weights should all be strictly positive.
"""
alns = get_alns_instance([lambda state, rnd: None],
[lambda state, rnd: None])
with assert_raises(ValueError):
alns.iterate(One(), [1, 1, 0, 1], .5, HillClimbing())
with assert_raises(ValueError):
alns.iterate(One(), [1, 1, -5, 1], .5, HillClimbing())
def test_raises_non_positive_iterations():
"""
The number of iterations should be strictly positive.
"""
alns = get_alns_instance([lambda state, rnd: None],
[lambda state, rnd: None])
with assert_raises(ValueError):
alns.iterate(One(), [1, 1, 1, 1], .5, HillClimbing(), 0)
with assert_raises(ValueError):
alns.iterate(One(), [1, 1, 1, 1], .5, HillClimbing(), -1)
def test_raises_boundary_operator_decay():
"""
The boundary cases, zero and one, should both raise.
"""
alns = get_alns_instance([lambda state, rnd: None],
[lambda state, rnd: None])
with assert_raises(ValueError):
alns.iterate(One(), [1, 1, 1, 1], 0, HillClimbing())
with assert_raises(ValueError):
alns.iterate(One(), [1, 1, 1, 1], 1, HillClimbing())
def test_raises_negative_iterations():
"""
The number of iterations should be non-negative, as zero is allowed.
"""
alns = get_alns_instance([lambda state, rnd: None],
[lambda state, rnd: None])
initial_solution = One()
# A negative iteration count is not understood, for obvious reasons.
with assert_raises(ValueError):
alns.iterate(initial_solution, [1, 1, 1, 1], .5, HillClimbing(), -1)
# But zero should just return the initial solution.
result = alns.iterate(initial_solution, [1, 1, 1, 1], .5, HillClimbing(), 0)
assert_(result.best_state is initial_solution)
def test_raises_missing_repair_operator():
"""
Tests if the algorithm raises when no repair operators have been set.
"""
alns = get_alns_instance(destroy_operators=[lambda state, rnd: None])
with assert_raises(ValueError):
alns.iterate(One(), [1, 1, 1, 1], 0.95, HillClimbing())
def test_does_not_raise():
"""
This set of parameters, on the other hand, should work correctly.
"""
alns = get_alns_instance([lambda state, rnd: One()],
[lambda state, rnd: One()])
alns.iterate(Zero(), [1, 1, 1, 1], .5, HillClimbing(), 100)
def test_raises_insufficient_weights():
"""
We need (at least) four weights to be passed-in, one for each updating
scenario.
"""
alns = get_alns_instance([lambda state, rnd: None],
[lambda state, rnd: None])
with assert_raises(ValueError):
alns.iterate(One(), [1, 1, 1], .5, HillClimbing())
def test_raises_explosive_operator_decay():
"""
Tests if the algorithm raises when an explosive operator decay parameter is
passed.
"""
alns = get_alns_instance([lambda state, rnd: None],
[lambda state, rnd: None])
with assert_raises(ValueError):
alns.iterate(One(), [1, 1, 1, 1], 1.2, HillClimbing())
def test_trivial_example():
"""
This tests the ALNS algorithm on a trivial example, where the initial
solution is zero, and any other operator returns one.
"""
alns = get_alns_instance([lambda state, rnd: Zero()],
[lambda state, rnd: Zero()])
result = alns.iterate(One(), [1, 1, 1, 1], .5, HillClimbing(), 100)
assert_equal(result.best_state.objective(), 0)
def test_on_best_is_called():
"""
Tests if the callback is invoked when a new global best is found.
"""
alns = get_alns_instance([lambda state, rnd: Zero()],
[lambda state, rnd: Zero()])
# Called when a new global best is found. In this case, that happens once:
# in the only iteration below. It returns a state with value 10, which
# should then also be returned by the entire algorithm.
alns.on_best(lambda *args: ValueState(10))
result = alns.iterate(One(), [1, 1, 1, 1], .5, HillClimbing(), 1)
assert_equal(result.best_state.objective(), 10)
def test_accepts_better():
"""
Tests if the hill climbing method accepts a better solution.
"""
hill_climbing = HillClimbing()
assert_(hill_climbing.accept(rnd.RandomState(), One(), One(), Zero()))
def test_rejects_worse():
"""
Tests if the hill climbing method accepts a worse solution.
"""
hill_climbing = HillClimbing()
assert_(not hill_climbing.accept(rnd.RandomState(), Zero(), Zero(), One()))
