from __future__ import annotations
from typing import TYPE_CHECKING
from gefest.core.geometry.datastructs.structure import Structure
if TYPE_CHECKING:
from gefest.core.configs.optimization_params import OptimizationParams
from golem.core.optimisers.meta.surrogate_optimizer import SurrogateEachNgenOptimizer
from golem.core.optimisers.objective import Objective
from gefest.core.opt.adapters.configuration_mapping import (
map_into_gpa,
map_into_graph_generation_params,
map_into_graph_requirements,
)
from gefest.tools.optimizers.optimizer import Optimizer
[docs]
class SurrogateOptimizer(Optimizer):
"""Wrapper for GOLEM SurrogateEachNgenOptimizer.
Provides optimization strategy using both physical simulator and surrogate model.
"""
def __init__(self, opt_params: OptimizationParams, initial_population=None, **kwargs) -> None:
super().__init__(opt_params.log_dispatcher, **kwargs)
self.opt_params = opt_params
self.objective = Objective(
quality_metrics={obj.__class__.__name__: obj for obj in opt_params.objectives},
is_multi_objective=len(opt_params.objectives) > 1,
)
self.requirements = map_into_graph_requirements(opt_params)
self.ggp = map_into_graph_generation_params(opt_params)
self.gpa = map_into_gpa(opt_params)
if initial_population:
self.initial_pop = initial_population
else:
self.initial_pop: list[Structure] = list(
map(opt_params.golem_adapter.adapt, opt_params.sampler(opt_params.pop_size)),
)
self.__surrogate_opt = SurrogateEachNgenOptimizer(
objective=self.objective,
initial_graphs=self.initial_pop,
requirements=self.requirements,
graph_generation_params=self.ggp,
graph_optimizer_params=self.gpa,
surrogate_each_n_gen=self.opt_params.golem_surrogate_each_n_gen,
)
[docs]
def optimize(self) -> list[Structure]:
"""Optimizes population.
Returns:
list[Structure]: Optimized population.
"""
optimized_graphs = self.__surrogate_opt.optimise(self.objective)
optimized_pop = list(map(self.opt_params.golem_adapter.restore, optimized_graphs))
return optimized_pop