A. N. Medakene, K. Bouanane
In academic conferences, the assignment of papers to reviewers can be formulated as a CO problem, named the Reviewer Assignment Problem (RAP). Given the matching degrees between reviewers and papers, we aim to find an assignment that maximizes the total matching degree such that the coverage and fairness constraints for each paper and balanced load constraint for each reviewer must be satisfied. As the fairness and load balance constraints are often neglected in existing works, we define the Balanced & Fair Reviewer Assignment Problem, BF-RAP, as a variant of RAP. It was shown in a previous work that BF-RAP can be reduced to Max m-ECP when we aim to find an equitable m-coloring in a defined graph such that the total performance of the partition is maximum. Where performance is defined for pairs of nonadjacent vertices. We present a new algorithm that aims to solve BF-RAP by finding an equitable m-coloring with maximum performance in the corresponding graph. After constructing an initial m-coloring with maximum performance, the algorithm aims to balance the color classes by moving vertices from overloaded to underloaded classes while minimizing the cost of such moves.
Keywords: Reviewer Assignment Problem, Fairness constraints, Load balance constraints, Equitable Coloring Problem.,
Scheduled
FB2 Heuristics 2
June 11, 2021 10:45 AM
2 - LV Kantorovich
