"""Self-adaptive cooperative enhanced scatter search (SACESS)."""
import itertools
import logging
import os
import time
from multiprocessing import Manager, Process
from multiprocessing.managers import SyncManager
from typing import Any, Dict, List, Optional, Tuple
import numpy as np
import pypesto
from ...startpoint import StartpointMethod
from ...store.read_from_hdf5 import read_result
from ...store.save_to_hdf5 import write_result
from ..optimize import Problem
from .ess import ESSExitFlag, ESSOptimizer
from .function_evaluator import FunctionEvaluatorMP, FunctionEvaluatorMT
from .refset import RefSet
__all__ = ["SacessOptimizer", "get_default_ess_options"]
logger = logging.getLogger(__name__)
[docs]class SacessOptimizer:
"""SACESS optimizer.
A shared-memory-based implementation of the SaCeSS algorithm presented in
[PenasGon2017]_. Multiple processes (`workers`) run consecutive ESSs in
parallel. After each ESS run, depending on the outcome, there is a chance
of exchanging good parameters, and changing ESS hyperparameters to those of
the most promising worker.
.. [PenasGon2017] 'Parameter estimation in large-scale systems biology models:
a parallel and self-adaptive cooperative strategy', David R. Penas,
Patricia González, Jose A. Egea, Ramón Doallo and Julio R. Banga,
BMC Bioinformatics 2017, 18, 52. https://doi.org/10.1186/s12859-016-1452-4
"""
[docs] def __init__(
self,
num_workers: Optional[int] = None,
ess_init_args: Optional[List[Dict[str, Any]]] = None,
max_walltime_s: float = np.inf,
sacess_loglevel: int = logging.INFO,
ess_loglevel: int = logging.WARNING,
):
"""Construct.
Parameters
----------
ess_init_args:
List of argument dictionaries passed to
:func:`ESSOptimizer.__init__`. Each entry corresponds to one worker
process. I.e., the length of this list is the number of ESSs.
Ideally, this list contains some more conservative and some more
aggressive configurations.
Resource limits such as ``max_eval`` apply to a single CESS
iteration, not to the full search.
Mutually exclusive with ``num_workers``.
num_workers:
Number of workers to be used. If this argument is given,
(different) default ESS settings will be used for each worker.
Mutually exclusive with ``ess_init_args``.
max_walltime_s:
Maximum walltime in seconds. Will only be checked between local
optimizations and other simulations, and thus, may be exceeded by
the duration of a local search. Defaults to no limit.
ess_loglevel:
Loglevel for ESS runs.
sacess_loglevel:
Loglevel for SACESS runs.
"""
if (num_workers is None and ess_init_args is None) or (
num_workers is not None and ess_init_args is not None
):
raise ValueError(
"Exactly one of `num_workers` or `ess_init_args` "
"has to be provided."
)
self.num_workers = num_workers or len(ess_init_args)
if self.num_workers < 2:
raise ValueError(
f"{self.__class__.__name__} needs at least 2 workers."
)
self.ess_init_args = ess_init_args
self.max_walltime_s = max_walltime_s
self.exit_flag = ESSExitFlag.DID_NOT_RUN
self.ess_loglevel = ess_loglevel
self.sacess_loglevel = sacess_loglevel
logger.setLevel(self.sacess_loglevel)
[docs] def minimize(
self,
problem: Problem,
startpoint_method: StartpointMethod,
):
"""Solve the given optimization problem."""
start_time = time.time()
logger.debug(
f"Running sacess with {self.num_workers} "
f"workers: {self.ess_init_args}"
)
ess_init_args = self.ess_init_args or get_default_ess_options(
num_workers=self.num_workers, dim=problem.dim
)
# shared memory manager to handle shared state
# (simulates the sacess manager process)
with Manager() as shmem_manager:
sacess_manager = SacessManager(
shmem_manager=shmem_manager,
ess_options=ess_init_args,
dim=problem.dim,
)
# create workers
workers = [
SacessWorker(
manager=sacess_manager,
ess_kwargs=ess_kwargs,
worker_idx=i,
max_walltime_s=self.max_walltime_s,
ess_loglevel=self.ess_loglevel,
)
for i, ess_kwargs in enumerate(ess_init_args)
]
# launch worker processes
worker_processes = [
Process(
name=f"sacess-worker-{i:02d}",
target=_run_worker,
args=(
worker,
problem,
startpoint_method,
self.sacess_loglevel,
),
)
for i, worker in enumerate(workers)
]
for p in worker_processes:
p.start()
# wait for finish
for p in worker_processes:
p.join()
walltime = time.time() - start_time
logger.info(
f"sacess stopping after {walltime:3g}s with global best "
f"{sacess_manager.get_best_solution()[1]}."
)
return self._create_result(problem)
def _create_result(self, problem: Problem) -> pypesto.Result:
"""Create result object.
Creates an overall Result object from the results saved by the workers.
"""
# gather results from workers and delete temporary result files
result = None
for worker_idx in range(self.num_workers):
tmp_result_filename = SacessWorker.get_temp_result_filename(
worker_idx
)
tmp_result = read_result(
tmp_result_filename, problem=False, optimize=True
)
os.remove(tmp_result_filename)
if result is None:
result = tmp_result
else:
result.optimize_result.append(
tmp_result.optimize_result,
sort=False,
prefix=f"{worker_idx}-",
)
result.optimize_result.sort()
result.problem = problem
return result
class SacessManager:
"""The Sacess manager.
Manages shared memory of a SACESS run. Loosely corresponds to the manager
process in [PenasGon2017]_.
Attributes
----------
_num_workers: Number of workers
_ess_options: ESS options for each worker
_best_known_fx: Best objective value encountered so far
_best_known_x: Parameters corresponding to ``_best_known_fx``
_worker_scores: Performance score of the different workers (the higher, the
more promising the respective worker is considered)
_worker_comms: Number of communications received from the individual
workers
_rejections: Number of rejected solutions received from workers since last
adaptation of ``_rejection_threshold``.
_rejection_threshold: Threshold for relative objective improvements that
incoming solutions have to pass to be accepted
_lock: Lock for accessing shared state.
_logger: A logger instance
"""
def __init__(
self,
shmem_manager: SyncManager,
ess_options: List[Dict[str, Any]],
dim: int,
):
self._num_workers = len(ess_options)
self._ess_options = [shmem_manager.dict(o) for o in ess_options]
self._best_known_fx = shmem_manager.Value("d", np.inf)
self._best_known_x = shmem_manager.Array("d", [np.nan] * dim)
self._rejections = shmem_manager.Value("i", 0)
# initial value from [PenasGon2017]_ p.9
self._rejection_threshold = shmem_manager.Value("d", 0.1)
# scores of the workers, ordered by worker-index
# initial score is the worker index
self._worker_scores = shmem_manager.Array(
"d", range(self._num_workers)
)
self._worker_comms = shmem_manager.Array("i", [0] * self._num_workers)
self._lock = shmem_manager.RLock()
self._logger = logging.getLogger()
def get_best_solution(self) -> Tuple[np.array, float]:
"""Get the best objective value and corresponding parameters."""
with self._lock:
return np.array(self._best_known_x), self._best_known_fx.value
def reconfigure_worker(self, worker_idx: int) -> Dict:
"""Reconfigure the given worker.
Updates the ESS options for the given worker to those of the worker at
the top of the scoreboard and returns those settings.
"""
with self._lock:
leader_options = self._ess_options[np.argmax(self._worker_scores)]
self._ess_options[worker_idx] = leader_options
return leader_options
def submit_solution(
self,
x: np.array,
fx: float,
sender_idx: int,
elapsed_time_s: float,
):
"""Submit a solution.
To be called by a worker.
Parameters
----------
x: Model parameters
fx: Objective value corresponding to ``x``
sender_idx: Index of the worker submitting the results.
elapsed_time_s: Elapsed time since the beginning of the sacess run.
"""
with self._lock:
# cooperation step
# solution improves best value by at least a factor of ...
if (
# initially _best_known_fx is NaN
(
np.isfinite(fx)
and not np.isfinite(self._best_known_fx.value)
)
# avoid division by 0. just accept any improvement if best
# known value is 0.
or (self._best_known_fx.value == 0 and fx < 0)
or (
fx < self._best_known_fx.value
and abs(
(self._best_known_fx.value - fx)
/ self._best_known_fx.value
)
> self._rejection_threshold.value
)
):
# accept solution
self._logger.debug(
f"Accepted solution from worker {sender_idx}: {fx}."
)
# accept
self._best_known_fx.value = fx
self._best_known_x.value = x
self._worker_comms[sender_idx] += 1
self._worker_scores[sender_idx] = (
self._worker_comms[sender_idx] * elapsed_time_s
)
else:
# reject solution
self._rejections.value += 1
self._logger.debug(
f"Rejected solution from worker {sender_idx} "
f"rel change: {abs((self._best_known_fx.value - fx) / self._best_known_fx.value)} "
f" < {self._rejection_threshold.value} "
f"(total rejections: {self._rejections.value})."
)
# adapt acceptance threshold if too many solutions have been
# rejected
if self._rejections.value > self._num_workers:
self._rejection_threshold.value /= 2
self._logger.debug(
"Lowered acceptance threshold to "
f"{self._rejection_threshold.value}."
)
self._rejections.value = 0
class SacessWorker:
"""A SACESS worker.
Repeatedly runs ESSs and exchanges information with a SacessManager.
Corresponds to a worker process in [PenasGon2017]_.
Attributes
----------
_manager: The sacess manager this worker is working for.
_worker_idx: Index of this worker.
_best_known_fx: Best objective value known to this worker (obtained on its
own or received from the manager).
_n_received_solutions: Number of solutions received by this worker since
the last one was sent to the manager.
_neval: Number of objective evaluations since the last solution was sent
to the manager.
_ess_kwargs: ESSOptimizer options for this worker (may get updated during
the self-adaptive step).
_acceptance_threshold: Minimum relative improvement of the objective
compared to the best known value to be eligible for submission to the
Manager.
_n_sent_solutions: Number of solutions sent to the Manager.
_max_walltime_s: Walltime limit.
_logger: A Logger instance.
_ess_loglevel: Logging level for ESS runs
"""
def __init__(
self,
manager: SacessManager,
ess_kwargs: Dict[str, Any],
worker_idx: int,
max_walltime_s: float = np.inf,
ess_loglevel: int = logging.WARNING,
):
self._manager = manager
self._worker_idx = worker_idx
self._best_known_fx = np.inf
self._n_received_solutions = 0
self._neval = 0
self._ess_kwargs = ess_kwargs
self._acceptance_threshold = 0.005
self._n_sent_solutions = 0
self._max_walltime_s = max_walltime_s
self._start_time = None
self._logger = logging.getLogger(str(os.getpid()))
# Set the manager logger to one created within the current process
self._manager._logger = self._logger
self._ess_loglevel = ess_loglevel
def run(
self,
problem: Problem,
startpoint_method: StartpointMethod,
):
"""Start the worker."""
self._logger.debug(
f"#{self._worker_idx} starting " f"({self._ess_kwargs})."
)
if n_procs := self._ess_kwargs.get('n_procs'):
evaluator = FunctionEvaluatorMP(
problem=problem,
startpoint_method=startpoint_method,
n_procs=n_procs,
)
else:
evaluator = FunctionEvaluatorMT(
problem=problem,
startpoint_method=startpoint_method,
n_threads=self._ess_kwargs.get('n_threads', 1),
)
self._start_time = time.time()
# create refset from ndiverse
refset = RefSet(
dim=self._ess_kwargs['dim_refset'], evaluator=evaluator
)
refset.initialize_random(
n_diverse=max(
self._ess_kwargs.get('n_diverse', 10 * problem.dim), refset.dim
)
)
i_iter = 0
ess_results = pypesto.Result(problem=problem)
while self._keep_going():
# run standard ESS
ess, cur_ess_results = self._run_ess(
refset=refset,
)
# drop all but the 50 best results
ess_results.optimize_result.append(
cur_ess_results.optimize_result,
prefix=f"{self._worker_idx}_{i_iter}_",
)
ess_results.optimize_result.list = (
ess_results.optimize_result.list[:50]
)
write_result(
ess_results,
self.get_temp_result_filename(self._worker_idx),
overwrite=True,
optimize=True,
)
# check if the best solution of the last local ESS is sufficiently
# better than the sacess-wide best solution
self.maybe_update_best(ess.x_best, ess.fx_best)
self._best_known_fx = min(ess.fx_best, self._best_known_fx)
# cooperation step
# try to obtain a new solution from manager
recv_x, recv_fx = self._manager.get_best_solution()
self._logger.log(
logging.DEBUG - 1,
f"Worker {self._worker_idx} received solution {recv_fx} "
f"(known best: {self._best_known_fx}).",
)
if recv_fx < self._best_known_fx or (
not np.isfinite(self._best_known_fx) and np.isfinite(recv_x)
):
self._logger.debug(
f"Worker {self._worker_idx} received better solution."
)
self._best_known_fx = recv_fx
self._n_received_solutions += 1
self.replace_solution(refset, x=recv_x, fx=recv_fx)
# Adaptive step
# Update ESS settings if we received way more solutions than we
# sent
# Magic numbers from [PenasGon2017]_ algorithm 5
if (
self._n_received_solutions > 10 * self._n_sent_solutions + 20
and self._neval > problem.dim * 5000
):
self._ess_kwargs = self._manager.reconfigure_worker(
self._worker_idx
)
self._logger.debug(
f"Updated settings on worker {self._worker_idx} to "
f"{self._ess_kwargs}"
)
self._logger.info(
f"sacess worker {self._worker_idx} iteration {i_iter} "
f"(best: {self._best_known_fx})."
)
i_iter += 1
def _run_ess(
self,
refset: RefSet,
) -> Tuple[ESSOptimizer, pypesto.Result]:
"""Run ESS."""
ess_kwargs = self._ess_kwargs.copy()
# account for sacess walltime limit
ess_kwargs['max_walltime_s'] = min(
ess_kwargs.get('max_walltime_s', np.inf),
self._max_walltime_s - (time.time() - self._start_time),
)
ess = ESSOptimizer(**ess_kwargs)
ess.logger.setLevel(self._ess_loglevel)
cur_ess_results = ess.minimize(
refset=refset,
)
self._logger.debug(
f"#{self._worker_idx}: ESS finished with best "
f"f(x)={ess.fx_best}"
)
return ess, cur_ess_results
def maybe_update_best(self, x: np.array, fx: float):
"""Maybe update the best known solution and send it to the manager."""
self._logger.debug(
f"Worker {self._worker_idx} maybe sending solution {fx}. "
f"best known: {self._best_known_fx}, "
f"rel change: {(self._best_known_fx - fx) / fx}, "
f"threshold: {self._acceptance_threshold}"
)
# solution improves best value by at least a factor of ...
if (
(np.isfinite(fx) and not np.isfinite(self._best_known_fx))
or (self._best_known_fx == 0 and fx < 0)
or (
fx < self._best_known_fx
and abs((self._best_known_fx - fx) / fx)
> self._acceptance_threshold
)
):
self._logger.debug(
f"Worker {self._worker_idx} sending solution {fx}."
)
self._n_sent_solutions += 1
self._best_known_fx = fx
# send best_known_fx to manager
self._neval = 0
self._n_received_solutions = 0
elapsed_time = time.time() - self._start_time
self._manager.submit_solution(
x=x,
fx=fx,
sender_idx=self._worker_idx,
elapsed_time_s=elapsed_time,
)
@staticmethod
def replace_solution(refset: RefSet, x: np.array, fx: float):
"""Replace the global refset member by the given solution."""
# [PenasGon2017]_ page 8, top
if "cooperative_solution" not in refset.attributes:
label = np.zeros(shape=refset.dim)
# on first call, mark the worst solution as "cooperative solution"
label[np.argmax(refset.fx)] = 1
refset.add_attribute("cooperative_solution", label)
# replace the cooperative solution
refset.update(
i=np.argwhere(refset.attributes["cooperative_solution"]),
x=x,
fx=fx,
)
def _keep_going(self):
"""Check exit criteria.
Returns
-------
``True`` if none of the exit criteria is met, ``False`` otherwise.
"""
# elapsed time
if time.time() - self._start_time >= self._max_walltime_s:
self.exit_flag = ESSExitFlag.MAX_TIME
self._logger.debug(
f"Max walltime ({self._max_walltime_s}s) exceeded."
)
return False
return True
@staticmethod
def get_temp_result_filename(worker_idx: int) -> str:
return f"sacess-{worker_idx:02d}_tmp.h5"
def _run_worker(
worker: SacessWorker,
problem: Problem,
startpoint_method: StartpointMethod,
sacess_loglevel: int,
):
"""Run the given SACESS worker.
Helper function as entrypoint for sacess worker processes.
"""
# different random seeds per process
np.random.seed((os.getpid() * int(time.time() * 1000)) % 2**32)
worker._logger.setLevel(sacess_loglevel)
return worker.run(problem=problem, startpoint_method=startpoint_method)
def get_default_ess_options(num_workers: int, dim: int) -> List[Dict]:
"""Get default ESS settings for (SA)CESS.
Returns settings for ``num_workers`` parallel scatter searches, combining
more aggressive and more conservative configurations.
Setting appropriate values for ``n_threads`` and ``local_optimizer`` is
left to the user. Defaults to single-threaded and no local optimizer.
Parameters
----------
num_workers: Number of configurations to return.
dim: Problem dimension.
"""
min_dimrefset = 3
settings = [
# settings for first worker
{
'dim_refset': 10 * dim,
'balance': 0.5,
'local_n2': 10,
},
# for the remaining workers, cycle through these settings
# 1
{
'dim_refset': max(min_dimrefset, dim),
'balance': 0.0,
'local_n1': 1,
'local_n2': 1,
},
# 2
{
'dim_refset': 3 * dim,
'balance': 0.0,
'local_n1': 4,
'local_n2': 4,
},
# 3
{
'dim_refset': 5 * dim,
'balance': 0.25,
'local_n1': 10,
'local_n2': 10,
},
# 4
{
'dim_refset': 10 * dim,
'balance': 0.5,
'local_n1': 20,
'local_n2': 20,
},
# 5
{
'dim_refset': 15 * dim,
'balance': 0.25,
'local_n1': 100,
'local_n2': 100,
},
# 6
{
'dim_refset': 12 * dim,
'balance': 0.25,
'local_n1': 50,
'local_n2': 50,
},
# 7
{
'dim_refset': int(7.5 * dim),
'balance': 0.25,
'local_n1': 15,
'local_n2': 15,
},
# 8
{
'dim_refset': 5 * dim,
'balance': 0.25,
'local_n1': 7,
'local_n2': 7,
},
# 9
{
'dim_refset': max(min_dimrefset, 2 * dim),
'balance': 0.0,
'local_n1': 2,
'local_n2': 2,
},
# 10
{
'dim_refset': max(min_dimrefset, int(0.5 * dim)),
'balance': 0.0,
'local_n1': 1,
'local_n2': 1,
},
]
return [
settings[0],
*(itertools.islice(itertools.cycle(settings[1:]), num_workers - 1)),
]