Method run

pyvrp/IteratedLocalSearch.py:85–197

Runs the iterated local search algorithm with the provided stopping criterion. The algorithm uses late acceptance hill-climbing as the acceptance criterion; see [1]_ for details. Parameters ---------- stop Stopping criterion to use. The a

(
        self,
        stop: StoppingCriterion,
        collect_stats: bool = True,
        display: bool = False,
        display_interval: float = 5.0,
    )

        self._params = params

    def run(
        self,
        stop: StoppingCriterion,
        collect_stats: bool = True,
        display: bool = False,
        display_interval: float = 5.0,
    ) -> Result:
        """
        Runs the iterated local search algorithm with the provided stopping
        criterion. The algorithm uses late acceptance hill-climbing as the
        acceptance criterion; see [1]_ for details.

        Parameters
        ----------
        stop
            Stopping criterion to use. The algorithm runs until the first time
            the stopping criterion returns ``True``.
        collect_stats
            Whether to collect statistics about the solver's progress. Default
            ``True``.
        display
            Whether to display information about the solver progress. Default
            ``False``. Progress information is only available when
            ``collect_stats`` is also set.
        display_interval
            Time (in seconds) between iteration logs. Defaults to 5s.

        Returns
        -------
        Result
            A Result object, containing statistics (if collected) and the best
            found solution.

        References
        ----------
        .. [1] Burke, E.K., and Y. Bykov (2017). The Late Acceptance
               Hill-Climbing Heuristic. *European Journal of Operational
               Research*, 258(1): 70 - 78.
               https://doi.org/10.1016/j.ejor.2016.07.012.
        """
        print_progress = ProgressPrinter(display, display_interval)
        print_progress.start(self._data)

        history: RingBuffer[Solution] = RingBuffer(self._params.history_length)
        stats = Statistics(collect_stats=collect_stats)

        start = time.perf_counter()
        iters = iters_no_improvement = 0
        best = curr = self._init

        cost_eval = self._pm.cost_evaluator()
        while not stop(cost_eval.cost(best)):
            iters += 1

            if iters_no_improvement == self._params.num_iters_no_improvement:
                print_progress.restart()
                history.clear()