Adaptive scheduling algorithm based on mixed graph model

This paper deals with an adaptive approach for scheduling problems, The main idea is to produce, for a class of similar problems, a special heuristic rule which is successful for problems of this class. It is realised by `tuning' the parameters of the algorithm while `learning' the peculiarities of considering a class of problems. Once trained on the sample problems (in the authors' experiments a well-known test problem with 10 jobs and 10 machines has been considered), the adaptive algorithm solves `close' problems better than a branch and bound algorithm with a time limit (both in running time and in accuracy of the constructed schedule) and better than the 25 heuristics (in accuracy) used in the learning stage. However, the adaptive algorithm is unable to perform better than the branch and bound algorithm with a time limit for such scheduling problems, which are not sufficiently close to the sample problem