def set_velocity(self, **kwargs):
if 'velocity' in kwargs:
self.velocity = kwargs.get('velocity')
if 'uppervs' in kwargs and 'lowervs' in kwargs:
self.velocity = helper.init_vector(self.n, kwargs['uppervs'],
kwargs['lowervs'])
def __init__(
self,
_np: int,
n: int,
ub, # upper boundary
lb, # low boundary
procedure: callable,
max_gen: int = 500,
optimal_minimal: bool = True,
boundary_strategy=Boundary.BOUNDARY,
random_state: int = None):
self.np = _np
self.n = n
self.ub = np.array(ub)
self.lb = np.array(lb)
self.procedure = procedure
self.max_gen = max_gen
self.optimal_minimal = optimal_minimal
# bs: Boundary Strategy
self.bs = Boundary.make_strategy(boundary_strategy)
self.random_state = random_state
if self.random_state:
np.random.seed(self.random_state)
# sc: Solution Collection
self.sc = np.array(
[init_vector(self.n, self.ub, self.lb) for _ in range(self.np)])
# fc: fitness Collection
self.fc = None
# hbsc: history best solution collection
self.hbsc = []
self.fsort = True
self.fappend = True
def create(self, class_str, all_zero=False):
# 对可行解进行初始化
vector = np.zeros(self.n) if all_zero else helper.init_vector(
self.n, self.us, self.ls)
_class = Solution
if class_str == 'TrialSolution':
_class = TrialSolution
if class_str == 'LabelSolution':
_class = LabelSolution
if class_str == 'VelocitySolution':
_class = VelocitySolution
return _class(vector)
def __init__(self, *args, vlb=None, vub=None, **kwargs):
super(PSO, self).__init__(*args, **kwargs)
self.vlb = vlb if vlb else ([-5] * self.n)
self.vub = vub if vub else ([5] * self.n)
self.vc = np.array([init_vector(self.n, self.vub, self.vlb) for _ in range(self.np)])
# self.vcg = lambda cg: self.vc - self.vc * cg / self.max_gen
# 记录每一个体的历史最优
self.psc = deepcopy(self.sc)
# 参数生成
self.wg = lambda cg: 0.9 - cg * 0.4 / self.max_gen
self.r1g = lambda: 0.05
self.r2g = lambda: 0.05
def create(n, us, ls):
return Solution(helper.init_vector(n, us, ls))
def scouter_stage(self, g):
for i, nc in enumerate(self.ncs.flatten().astype(int)):
if nc >= self.nc:
self.sc[i] = init_vector(self.n, self.ub, self.lb)
self.ncs[i] = 0