Companion to Off-line and On-line Tuning: A Study on Operator Selection for a Memetic Algorithm Applied to the QAP

In the paper [1] we study the performance of three memetic algorithms when their configuration is tuned either off-line or on-line. We assess the relative performance achieved by algorithms of different quality tuned either off-line or on-line. We tackle the quadratic assignment problem. The results suggest that off-line tuning achieves in general better performance than on-line tuning. On-line is either comparable or more profitable than off-line tuning only if the quality of the algorithm is rather poor. Here we present the results of the whole experimental analysis, on three memetic algoritms (MA) and on two MAX–MIN ant system (MMAS) .