|
Last change
on this file since 1163 was
1145,
checked in by Maciej Komosinski, 3 years ago
|
|
Added niching and novelty search with limited (i.e., local) competition ("nearest neighbors" according to dissimilarity measure)
|
|
File size:
1.2 KB
|
| Line | |
|---|
| 1 | from abc import ABC |
|---|
| 2 | |
|---|
| 3 | from evolalg.base.step import Step |
|---|
| 4 | import numpy as np |
|---|
| 5 | |
|---|
| 6 | |
|---|
| 7 | class Dissimilarity(Step, ABC): |
|---|
| 8 | |
|---|
| 9 | def __init__(self, reduction="mean", output_field="dissim", knn=None, *args, **kwargs): |
|---|
| 10 | super(Dissimilarity, self).__init__(*args, **kwargs) |
|---|
| 11 | |
|---|
| 12 | self.output_field = output_field |
|---|
| 13 | self.fn_reduce = None |
|---|
| 14 | self.knn = knn |
|---|
| 15 | if reduction == "mean": # TODO change this 'elif' sequence to dictionary? |
|---|
| 16 | self.fn_reduce = np.mean |
|---|
| 17 | elif reduction == "max": |
|---|
| 18 | self.fn_reduce = np.max |
|---|
| 19 | elif reduction == "min": |
|---|
| 20 | self.fn_reduce = np.min |
|---|
| 21 | elif reduction == "sum": |
|---|
| 22 | self.fn_reduce = np.sum |
|---|
| 23 | elif reduction == "knn_mean": |
|---|
| 24 | self.fn_reduce = self.knn_mean |
|---|
| 25 | elif reduction == "none" or reduction is None: |
|---|
| 26 | self.fn_reduce = None |
|---|
| 27 | else: |
|---|
| 28 | raise ValueError("Unknown reduction type. Supported: mean, max, min, sum, knn_mean, none") |
|---|
| 29 | |
|---|
| 30 | def reduce(self, dissim_matrix): |
|---|
| 31 | if self.fn_reduce is None: |
|---|
| 32 | return dissim_matrix |
|---|
| 33 | return self.fn_reduce(dissim_matrix, axis=1) |
|---|
| 34 | |
|---|
| 35 | def knn_mean(self, dissim_matrix,axis): |
|---|
| 36 | return np.mean(np.partition(dissim_matrix, self.knn)[:,:self.knn],axis=axis) |
|---|
Note: See
TracBrowser
for help on using the repository browser.
