Cluster approach to the domains formation
Leonid B. Litinskii · Optical Memory and Neural Networks · 2007
As a rule, the quadratic functional depending on a great number of binary variables has a lot of local minima. One of approaches allowing to find (in averaged) deeper local minima is aggregation of binary variables into larger blocks/domains. To minimize the functional one has to change the states of aggregated variables (domains). In present publication we discuss methods of domains formation. It is shown that the best results are obtained when domains are formed by variables that are strongly connected with each other.