[BUG] Incorrect strength calculation axis in SPEA2Survival._do method #748
Description
Checklist
- I searched existing issues
- I'm using the latest pymoo version
Bug Description
In the SPEA2Survival._do method, the calculation of the strength S uses S = (M == 1).sum(axis=0), which counts how many times each individual is dominated by others, instead of how many individuals each dominates. According to the SPEA2 algorithm, S should be computed as S = (M == 1).sum(axis=1) (i.e., summing across rows), so S[i] is the number of solutions dominated by individual i. This could lead to incorrect fitness calculations and selection in the algorithm.
Minimal Code to Reproduce
# Example code snippet for reproduction:
import numpy as np
from pymoo.util.dominator import Dominator
F = np.array([[1,2],[2,3],[3,4]])
M = Dominator().calc_domination_matrix(F)
S_col = (M == 1).sum(axis=0) # current implementation
S_row = (M == 1).sum(axis=1) # correct per SPEA2
print('Current S:', S_col)
print('Correct S:', S_row)Error Message
Current S: [0 1 2]
Correct S: [2 1 0]
The current S assigns the strength as the number of times an individual is dominated, rather than how many it dominates. This contradicts the original SPEA2 definition and may affect algorithm behavior.PyMoo & Python Version
Pymoo 0.6.1.5, Python 3.13