Self-adaptive Differential Evolution with Neighborhood Search

Notes

DE

operators:
- mutation:
  - (1) $v_i = x_{i_1} + F \cdot (x_{i_2} - x_{i_3})$ ;
  - (2) $v_i = x_b + F \cdot (x_{i_1} - x_{i_2})$ ;
  - (3) $v_i = x_i + F \cdot (x_b - x_i) + F \cdot (x_{i_1} - x_{i_2})$ ;
  - (4) $v_i = x_b + F \cdot (x_{i_1} - x_{i_2}) + F \cdot (x_{i_3} - x_{i_4})$ ;
  - (5) $v_i = x_{i_1} + F \cdot (x_{i_1} - x_{i_2}) + F \cdot (x_{i_3} - x_{i_4})$ ;

SaDE

operators:
- mutation:
  - ;
  - adaptation:
    - , : the number of offsprings (by using (1) or (3) strategy) that are moved to the new generation;
    - , : the same as above but the rejected offsprings;
    - the previous numbers are collected for a number of generations (learning period) (they've used 50 for SaDE);
    - $p = \frac{ ns_1 \cdot (ns_2 + ns_3)}{ ns_2 \cdot (ns_1 + nf_1) + ns_1 \cdot (ns_2 + nf_2)}$ ;
  - ;
- crossover:
  - adaptation:
    - for each phenotype a parameter is generated;
    - the generated parameter is preserved for a couple of generations (they've used 5 for SaDE);
    - the generated parameters that induce successful offsprings are collected for a number of generations (they've used 25 for SaDE); after which a new is generated;
    - ;
    - ;