Fix distance calcuation in NSGA-II #86
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…pdated test to be more robust to shuffled input
danyoungday
commented
May 20, 2024
| @@ -59,20 +59,24 @@ def fast_non_dominated_sort(candidates: list): | |||
|
|
|||
| def calculate_crowding_distance(front): | |||
| """ | |||
| Calculate crowding distance of each candidate in front. | |||
| Set crowding distance of each candidate in front. | |||
There was a problem hiding this comment.
Change from creating a list of distances to setting the distances ourselves.
danyoungday
commented
May 20, 2024
| distances[i] += (front[i+1].metrics[m] - front[i-1].metrics[m]) / (obj_max - obj_min) | ||
|
|
||
| return distances | ||
| sorted_front = sorted(front, key=lambda candidate: candidate.metrics[m]) |
There was a problem hiding this comment.
No longer sort in-place so we don't mess anything up outside this function
danyoungday
commented
May 20, 2024
| for i in range(1, len(sorted_front) - 1): | ||
| if obj_max != obj_min: | ||
| dist = sorted_front[i+1].metrics[m] - sorted_front[i-1].metrics[m] | ||
| sorted_front[i].distance += dist / (obj_max - obj_min) |
There was a problem hiding this comment.
Set distance rather than creating distance list so that we ensure we are setting the correct distance to the correct value
danyoungday
commented
May 20, 2024
| crowding_distance = nsga2_utils.calculate_crowding_distance(front) | ||
| for candidate, distance in zip(front, crowding_distance): | ||
| candidate.distance = distance | ||
| nsga2_utils.calculate_crowding_distance(front) |
There was a problem hiding this comment.
Change from manually setting distances outside function to changing them inplace
danyoungday
commented
May 20, 2024
| nsga2_utils.calculate_crowding_distance(shuffled_front) | ||
| for candidate, tgt in zip(shuffled_front, shuffled_tgts): | ||
| self.assertAlmostEqual(candidate.distance, tgt) | ||
|
|
There was a problem hiding this comment.
Tests a simple example. Makes sure the front is shuffled so that we can see the behavior works in a real scenario
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Distance calculation was being done wrong before! Was being done randomly since we weren't sorting the distance list. Copied distance calculation from merging experiment to work properly. We were still getting a good pareto because we had so many candidates and generations, but now we may do better than before or converge much faster.