def test_update_gbest_neighborhood(swarm, p, k, static):
"""Test if update_gbest_neighborhood gives the expected return values"""
topology = Ring(static=static)
pos, cost = topology.compute_gbest(swarm, p=p, k=k)
expected_pos = np.array([9.90438476e-01, 2.50379538e-03, 1.87405987e-05])
expected_cost = 1.0002528364353296
assert cost == pytest.approx(expected_cost)
assert pos == pytest.approx(expected_pos)
def test_update_gbest_neighborhood(swarm, p, k):
"""Test if update_gbest_neighborhood gives the expected return values"""
topology = Ring()
pos, cost = topology.compute_gbest(swarm, p=p, k=k)
expected_pos = np.array([1,2,3])
expected_cost = 1
assert (pos == expected_pos).all()
assert cost == expected_cost
def __fitNBeta(self, dim, n_particles, itera, options, objetive_function,
BetaChange, bound):
my_topology = Ring()
my_swarm = P.create_swarm(n_particles=n_particles,
dimensions=dim,
options=options,
bounds=bound)
my_swarm.pbest_cost = np.full(n_particles, np.inf)
my_swarm.best_cost = np.inf
for i in range(itera):
for a in range(n_particles):
my_swarm.position[a][0:BetaChange] = sorted(
my_swarm.position[a][0:BetaChange])
for c in range(1, self.BetaChange):
if my_swarm.position[a][c -
1] + 5 >= my_swarm.position[a][c]:
my_swarm.position[a][c] = my_swarm.position[a][c] + 5
my_swarm.current_cost = objetive_function(my_swarm.position)
my_swarm.pbest_pos, my_swarm.pbest_cost = P.operators.compute_pbest(
my_swarm)
#my_swarm.current_cost[np.isnan(my_swarm.current_cost)]=np.nanmax(my_swarm.current_cost)
#my_swarm.pbest_cost = objetive_function(my_swarm.pbest_pos)
my_swarm.best_pos, my_swarm.best_cost = my_topology.compute_gbest(
my_swarm, options['p'], options['k'])
if i % 20 == 0:
print(
'Iteration: {} | my_swarm.best_cost: {:.4f} | days: {}'.
format(
i + 1, my_swarm.best_cost,
str(my_swarm.pbest_pos[my_swarm.pbest_cost.argmin()])))
my_swarm.velocity = my_topology.compute_velocity(my_swarm,
bounds=bound)
my_swarm.position = my_topology.compute_position(my_swarm,
bounds=bound)
final_best_cost = my_swarm.best_cost.copy()
final_best_pos = my_swarm.pbest_pos[
my_swarm.pbest_cost.argmin()].copy()
return final_best_pos, final_best_cost
def test_neighbor_idx(swarm, static, p, k):
"""Test if the neighbor_idx attribute is assigned"""
topology = Ring(static=static)
topology.compute_gbest(swarm, p=p, k=k)
assert topology.neighbor_idx is not None
class DragonFlyOptimizer(DiscreteSwarmOptimizer):
def assertions(self):
if self.velocity_clamp is not None:
if not isinstance(self.velocity_clamp, tuple):
raise TypeError("Parameter `velocity_clamp` must be a tuple")
if not len(self.velocity_clamp) == 2:
raise IndexError("Parameter `velocity_clamp` must be of "
"size 2")
if not self.velocity_clamp[0] < self.velocity_clamp[1]:
raise ValueError("Make sure that velocity_clamp is in the "
"form (v_min, v_max)")
def __init__(
self,
n_particles,
dimensions,
options={},
init_pos=None,
velocity_clamp=None,
ftol=-np.inf,
):
self.logger = logging.getLogger(__name__)
super().__init__(
n_particles=n_particles,
dimensions=dimensions,
options=options,
binary=True,
init_pos=init_pos,
velocity_clamp=velocity_clamp,
ftol=ftol,
)
self.food_fitness = np.inf
self.enemy_fitness = -np.inf
self.food_pos = np.empty(0)
self.enemy_pos = np.empty(0)
self.assertions()
# Initialize the resettable attributes
self.reset()
# Initialize the topology
self.top = Ring(static=False)
def compute_pworst(self, swarm): # Compute enemy position and cost
try:
# Infer dimensions from positions
dimensions = swarm.dimensions
# Create a 1-D and 2-D mask based from comparisons
mask_cost = swarm.current_cost > swarm.pbest_cost
mask_pos = np.repeat(mask_cost[:, np.newaxis], dimensions, axis=1)
# Apply masks
new_pworst_pos = np.where(~mask_pos, swarm.pbest_pos,
swarm.position)
new_pworst_cost = np.where(~mask_cost, swarm.pbest_cost,
swarm.current_cost)
except AttributeError:
msg = "Please pass a Swarm class. You passed {}".format(
type(swarm))
self.logger.error(msg)
raise
else:
return (new_pworst_pos, new_pworst_cost)
def _transfer_function(self, v):
"""Helper method for the transfer function
Parameters
----------
x : numpy.ndarray
Input vector for sigmoid computation
Returns
-------
numpy.ndarray
Output transfer function computation
"""
return abs(v / np.sqrt(v**2 + 1))
def compute_position(self, velocity):
return np.random.random_sample(
size=self.dimensions) < self._transfer_function(velocity)
def optimize(self,
objective_func,
iters,
print_step=1,
verbose=1,
**kwargs):
ub = 1
lb = 0
for i in range(iters):
w = 0.9 - i * ((0.9 - 0.4) / iters)
my_c = 0.1 - i * ((0.1 - 0) / (iters / 2))
if my_c < 0:
my_c = 0
# print(my_c)
s = 2 * random.random() * my_c # Seperation weight
a = 2 * random.random() * my_c # Alignment weight
c = 2 * random.random() * my_c # Cohesion weight
f = 2 * random.random() # Food attraction weight
e = my_c # Enemy distraction weight
# Compute cost for current position and personal best
self.swarm.current_cost = objective_func(self.swarm.position,
**kwargs)
self.swarm.pbest_cost = objective_func(self.swarm.pbest_pos,
**kwargs)
self.swarm.pbest_pos, self.swarm.pbest_cost = compute_pbest(
self.swarm)
self.swarm.pworst_pos, self.swarm.pworst_cost = self.compute_pworst(
self.swarm)
pmin_cost_idx = np.argmin(self.swarm.pbest_cost)
pmax_cost_idx = np.argmax(self.swarm.pworst_cost)
# pmax_cost_idx = np.argmax(self.swarm.pbest_cost)
# Update gbest from neighborhood
# self.swarm.best_cost = np.min(self.swarm.pbest_cost)
# self.swarm.pbest_pos = self.swarm.pbest_pos[np.argmin(self.swarm.pbest_cost)]
# best_cost_yet_found = np.min(self.swarm.best_cost)
self.swarm.best_pos, self.swarm.best_cost = self.top.compute_gbest(
self.swarm, 2, self.n_particles)
# Updating Food position
if self.swarm.pbest_cost[pmin_cost_idx] < self.food_fitness:
self.food_fitness = self.swarm.pbest_cost[pmin_cost_idx]
self.food_pos = self.swarm.pbest_pos[pmin_cost_idx]
# Updating Enemy position
if self.swarm.pworst_cost[pmax_cost_idx] > self.enemy_fitness:
self.enemy_fitness = self.swarm.pworst_cost[pmax_cost_idx]
self.enemy_pos = self.swarm.pworst_pos[pmax_cost_idx]
# best_cost_yet_found = np.min(self.swarm.best_cost)
for j in range(self.n_particles):
S = np.zeros(self.dimensions)
A = np.zeros(self.dimensions)
C = np.zeros(self.dimensions)
F = np.zeros(self.dimensions)
E = np.zeros(self.dimensions)
# Calculating Separation(S)
for k in range(self.n_particles):
S += (self.swarm.position[k] - self.swarm.position[j])
S = -S
# Calculating Allignment(A)
for k in range(self.n_particles):
A += self.swarm.velocity[k]
A = (A / self.n_particles)
# Calculating Cohesion
for k in range(self.n_particles):
C += self.swarm.position[k]
C = (C / self.n_particles) - self.swarm.position[j]
F = self.food_pos - self.swarm.position[
j] # Calculating Food postion
E = self.enemy_pos - self.swarm.position[
j] # Calculating Enemy position
self.swarm.velocity[j] = (s * S + a * A + c * C + f * F +
e * E) + w * self.swarm.velocity[j]
self.swarm.position[j] = self.compute_position(
self.swarm.velocity[j])
# Print to console
if i % print_step == 0:
cli_print(
"Iteration {}/{}, cost: {}".format(
i + 1, iters, np.min(self.swarm.best_cost)),
verbose,
2,
logger=self.logger,
)
# Obtain the final best_cost and the final best_position
# final_best_cost = np.min(self.swarm.pbest_cost)
# final_best_pos = self.swarm.pbest_pos[np.argmin(self.swarm.pbest_cost)]
final_best_cost = self.swarm.best_cost.copy()
final_best_pos = self.swarm.best_pos.copy()
print("==============================\nOptimization finished\n")
print("Final Best Cost : ", final_best_cost, "\nBest Value : ",
final_best_pos)
# end_report(
# final_best_cost, final_best_pos, verbose, logger=self.logger
# )
return (final_best_cost, final_best_pos)
class DiscreteBoundedPSO(BinaryPSO):
"""
This class is based on the Binary PSO class. It extends the BinaryPSO class
by a function which allows the conversion of discrete optimization variables
into binary variables, so that discrete optimization problems can be solved
"""
def __init__(
self,
n_particles,
dimensions_discrete,
options,
bounds,
bh_strategy="periodic",
init_pos=None,
velocity_clamp=None,
vh_strategy="unmodified",
ftol=-np.inf,
ftol_iter=1,
):
"""Initialize the swarm
Attributes
----------
n_particles : int
number of particles in the swarm.
dimensions_discrete : int
number of discrete dimensions of the search space.
options : dict with keys :code:`{'c1', 'c2', 'w', 'k', 'p'}`
a dictionary containing the parameters for the specific
optimization technique
* c1 : float
cognitive parameter
* c2 : float
social parameter
* w : float
inertia parameter
* k : int
number of neighbors to be considered. Must be a
positive integer less than :code:`n_particles`
* p: int {1,2}
the Minkowski p-norm to use. 1 is the
sum-of-absolute values (or L1 distance) while 2 is
the Euclidean (or L2) distance.
bounds : tuple of numpy.ndarray
a tuple of size 2 where the first entry is the minimum bound while
the second entry is the maximum bound. Each array must be of shape
:code:`(dimensions,)`.
init_pos : numpy.ndarray, optional
option to explicitly set the particles' initial positions. Set to
:code:`None` if you wish to generate the particles randomly.
velocity_clamp : tuple, optional
a tuple of size 2 where the first entry is the minimum velocity
and the second entry is the maximum velocity. It
sets the limits for velocity clamping.
vh_strategy : String
a strategy for the handling of the velocity of out-of-bounds particles.
Only the "unmodified" and the "adjust" strategies are allowed.
ftol : float
relative error in objective_func(best_pos) acceptable for
convergence
ftol_iter : int
number of iterations over which the relative error in
objective_func(best_pos) is acceptable for convergence.
Default is :code:`1`
"""
# Initialize logger
self.rep = Reporter(logger=logging.getLogger(__name__))
# Assign k-neighbors and p-value as attributes
self.k, self.p = options["k"], options["p"]
self.dimensions_discrete = dimensions_discrete
self.bits, self.bounds = self.discretePSO_to_binaryPSO(
dimensions_discrete, bounds)
# Initialize parent class
super(BinaryPSO, self).__init__(
n_particles=n_particles,
dimensions=sum(self.bits),
binary=True,
options=options,
init_pos=init_pos,
velocity_clamp=velocity_clamp,
ftol=ftol,
ftol_iter=ftol_iter,
)
# self.bounds = bounds
# Initialize the resettable attributes
self.reset()
# Initialize the topology
self.top = Ring(static=False)
self.vh = VelocityHandler(strategy=vh_strategy)
self.bh = BoundaryHandler(strategy=bh_strategy)
self.name = __name__
def optimize(self,
objective_func,
iters,
n_processes=None,
verbose=True,
**kwargs):
"""Optimize the swarm for a number of iterations
Performs the optimization to evaluate the objective
function :code:`f` for a number of iterations :code:`iter.`
Parameters
----------
objective_func : function
objective function to be evaluated
iters : int
number of iterations
n_processes : int, optional
number of processes to use for parallel particle evaluation
Defaut is None with no parallelization.
verbose : bool
enable or disable the logs and progress bar (default: True = enable logs)
kwargs : dict
arguments for objective function
Returns
-------
tuple
the local best cost and the local best position among the
swarm.
"""
# Apply verbosity
if verbose:
log_level = logging.INFO
else:
log_level = logging.NOTSET
self.rep.log("Obj. func. args: {}".format(kwargs), lvl=logging.DEBUG)
self.rep.log(
"Optimize for {} iters with {}".format(iters, self.options),
lvl=log_level,
)
# Populate memory of the handlers
self.bh.memory = self.swarm.position
self.vh.memory = self.swarm.position
# Setup Pool of processes for parallel evaluation
pool = None if n_processes is None else mp.Pool(n_processes)
self.swarm.pbest_cost = np.full(self.swarm_size[0], np.inf)
ftol_history = deque(maxlen=self.ftol_iter)
for i in self.rep.pbar(iters, self.name) if verbose else range(iters):
# Compute cost for current position and personal best
''' Binary swarm postitions need to be transformed to discrete swarm
postitions first, because the objective function expects discrete
values (only positions are transformed!), original binary
position is saved in binary_swarm_position'''
binary_swarm_position = self.BinarySwarmPositions_to_DiscreteSwarmPositions(
)
# Evaluate Cost Function
self.swarm.current_cost = compute_objective_function(
self.swarm, objective_func, pool, **kwargs)
''' Transform discrete swarm positions back to binary positions,
because the PSO works on binary particles'''
self.swarm.position = binary_swarm_position
self.swarm.pbest_pos, self.swarm.pbest_cost = compute_pbest(
self.swarm)
best_cost_yet_found = np.min(self.swarm.best_cost)
# Update gbest from neighborhood
self.swarm.best_pos, self.swarm.best_cost = self.top.compute_gbest(
self.swarm, p=self.p, k=self.k)
if verbose:
# Print to console
self.rep.hook(best_cost=self.swarm.best_cost)
# Save to history
hist = self.ToHistory(
best_cost=self.swarm.best_cost,
mean_pbest_cost=np.mean(self.swarm.pbest_cost),
mean_neighbor_cost=np.mean(self.swarm.best_cost),
position=self.swarm.position,
velocity=self.swarm.velocity,
)
self._populate_history(hist)
# Verify stop criteria based on the relative acceptable cost ftol
relative_measure = self.ftol * (1 + np.abs(best_cost_yet_found))
delta = (np.abs(self.swarm.best_cost - best_cost_yet_found) <
relative_measure)
if i < self.ftol_iter:
ftol_history.append(delta)
else:
ftol_history.append(delta)
if all(ftol_history):
break
# Perform position velocity update
self.swarm.velocity = self.top.compute_velocity(
self.swarm, self.velocity_clamp, self.vh)
self.swarm.position = self._compute_position(self.swarm)
# Obtain the final best_cost and the final best_position
final_best_cost = self.swarm.best_cost.copy()
final_best_pos = self.swarm.pbest_pos[
self.swarm.pbest_cost.argmin()].copy()
self.rep.log(
"Optimization finished | best cost: {}, best pos: {}".format(
final_best_cost, final_best_pos),
lvl=log_level,
)
# Close Pool of Processes
if n_processes is not None:
pool.close()
return (final_best_cost, final_best_pos)
def discretePSO_to_binaryPSO(self, dimensions_discrete, bounds):
"""
Translate a discrete PSO-problem into a binary PSO-problem by
calculating the number of bits necessary to represent the discrete
optimization problem with "dimensions_discrete" number of discrete
variables as a binary optimization problem. The bounds are encoded in
the binary representation and might be tightened.
Parameters
----------
dimensions_discrete: integer
dimension of the discrete search space.
bounds : tuple of numpy.ndarray
a tuple of size 2 where the first entry is the minimum bound while
the second entry is the maximum bound. Each array must be of shape
:code:`(dimensions,)`.
"""
bits = []
for n in range(0, dimensions_discrete):
# Number of bits required rounding down!
bits.append(
int(np.log10(bounds[1][n] - bounds[0][n] + 1) / np.log10(2)))
# Adjust upper bound accordingly
bounds[1][n] = bounds[0][n] + 2**bits[n] - 1
return bits, bounds
def BinarySwarmPositions_to_DiscreteSwarmPositions(self):
"""
Converts binary self.swarm.position to discrete values. Returns the
original binary position, so that it can be used to restore
self.swarm.position to the original binary values.
"""
binary_position = self.swarm.position
discrete_position = np.zeros(
(self.n_particles, self.dimensions_discrete))
cum_sum = 0
for i in range(0, self.dimensions_discrete):
bit = self.bits[i]
lb = self.bounds[0][i]
discrete_position[:,[i]] = lb + \
self.bool2int(binary_position[:,cum_sum:cum_sum+bit])
cum_sum = cum_sum + bit
# Set swarm position to discrete integer values
self.swarm.position = discrete_position.astype(int)
return binary_position
def bool2int(self, x):
"""
Converts a binary variable represented by an array x (row vector) into
an integer value
"""
x_int = np.zeros((x.shape[0], 1))
for row in range(0, x.shape[0]):
row_int = 0
for i, j in enumerate(x[row, :]):
row_int += j << i
x_int[row] = row_int
return x_int
