We compare two alternative ways to use the Bayesian approach in heuristic optimization. The “no-learning” way means that we optimize the randomization parameters for each problem separately. The “learning” way means that we optimize the randomization parameters for some “learning” set of problems. We use those parameters later on for a family of related problems.

We define the learning efficiency as a non-uniformity of optimal parameters while solving a set of randomly generated problems. We show that for flow-shop problems the non-uniformity of optimal parameters is significant. It means that the Bayesian learning is efficient in those problems.