An efficient algorithm is given for finding global minimum points of Hopfield-type energy functions based on a partitioning technique. At the first step, the energy function is partitioned into small systems, and all the local minimum points are found for each partitioned function, some of them being chosen as candidates for the global minimum points. At the second step, the energy of the original function is estimated at the points obtained by combinations of these candidates from the partitioned functions.
展开▼
