protected Solution makeHeuristicMove(
Deadline deadline, Solution s, Random random, HeuristicStats heuristicStats) {
Heuristic heuristic = heuristicStats.getHeuristic();
long costBefore = s.getCost();
long startTime = System.currentTimeMillis();
s =
heuristic.move(
s,
Deadline.min(new Deadline(heuristicMaxTimes[heuristicStats.getId()]), deadline),
random);
long timeDiff = System.currentTimeMillis() - startTime;
long costDiff = costBefore - s.getCost();
heuristicStats.saveUsage(timeDiff, costDiff);
logger.info(
String.format(
"Heuristic # %d has improved the solution by %d (%f %%) in %d ms; its last performance = %f",
heuristicStats.getId(),
costDiff,
(double) costDiff * 100 / costBefore,
timeDiff,
heuristicStats.getLastPerformance()));
return s;
}
@Override
public Solution solve(Problem problem, Solution initialSolution, Deadline deadline) {
exploitedHeuristics.clear();
heuristicStats.clear();
queue.clear();
Solution s = initialSolution;
if (!(initialSolution instanceof SmartSolution)) {
s = new SmartSolution(s);
}
Random random = problem.getRandom();
for (int i = 0; i < heuristics.size(); i++) {
HeuristicStats stats =
new HeuristicStats(heuristics.get(i), i, useLastPerformanceOnly, saveZeroImprovement);
heuristicStats.add(stats);
s = processHeuristicMove(deadline, s, random, stats);
}
while (!deadline.hasExpired() && queue.size() > 0) {
logger.info("QUEUE State:");
int counter = 0;
toRecompute.clear();
long now = System.currentTimeMillis();
for (HeuristicStats stats : queue) {
if (now - stats.getLastUseTime() > recomputeTime) {
logger.info(
"RECOMPUTING: Heuristic # "
+ stats.getId()
+ " was not applied for "
+ (now - stats.getLastUseTime())
+ " ms");
toRecompute.add(stats);
}
logger.info(
counter + " : heuristic # " + stats.getId() + " -> " + stats.getLastPerformance());
counter++;
}
for (HeuristicStats stats : toRecompute) {
s = processHeuristicMove(deadline, s, random, stats);
Safety.saveSolution(s);
}
HeuristicStats bestHeuristic = queue.pollFirst();
s = processHeuristicMove(deadline, s, random, bestHeuristic);
Safety.saveSolution(s);
}
for (HeuristicStats stats : heuristicStats) {
stats.logPerformance();
}
return s;
}
private Solution processHeuristicMove(
Deadline deadline, Solution s, Random random, HeuristicStats heuristicStats) {
Heuristic heuristic = heuristicStats.getHeuristic();
double costBefore = s.getCost();
s = makeHeuristicMove(deadline, s, random, heuristicStats);
double costDiff = costBefore - s.getCost();
if (costDiff > 0) {
queue.add(heuristicStats);
for (HeuristicStats stats : exploitedHeuristics) {
queue.add(stats);
}
exploitedHeuristics.clear();
} else if (heuristic.isDeterministic()) {
exploitedHeuristics.add(heuristicStats);
} else {
queue.add(heuristicStats);
}
return s;
}
