Doi: 10. 1016/S0305-0548(03)00202-8
Download 490.36 Kb. Pdf ko'rish
|
10.1016@S0305-05480300202-8
r
level j level j+1 level j-1 Fig. 4. Algorithm illustration: before (left) and after (right) Recluster. we create trees from these star-like clusters to evaluate the actual solution. Operation Tree is also used in the TreePlan algorithm for creating trees under the already planned parts. 5.1.1.3. Moves. This operation performs local optimization on a candidate solution in order to improve it. Operation Moves detaches a node from its original parent node and connects it to a new parent node. The move is accepted if it yields a lower solution cost. Two subsets of N are the input for this operation. The /rst set C includes the candidate nodes that can be moved, while the second one P contains the parent nodes to be examined as target nodes for a move. We try all candidate–parent pairs for possible moves while they improve the current solution. This operation is used by compound operations Recluster and LocalOpt. 5.1.1.4. Swaps. This operation also performs local optimization on a given solution. Here the elementary step interchanges (swaps) the parent nodes of two given candidate nodes. We accept a swap only if it results lower cost. The input of Swaps is a subset C ⊆ N containing the candidate nodes that are allowed to be swapped. We try all possible pairs of candidate nodes for swapping while they improve the current solution. This operation is used by compound operations Recluster and LocalOpt. 5.1.2. Compound operations 5.1.2.1. Recluster. This operation recalculates the number and positions of the concentrators in a selected part of the network and integrates operation Cluster with local optimization operations Moves and Swaps. It has three input parameters: a node r, which de/nes the subsection of the solution to be reorganized, the type of allocation step within Cluster and the frequency of local optimization steps. Operation Recluster works as follows (see Fig. 4 ): (1) Consider the child nodes of r in level j and their children in level j + 1. (2) Select new nodes to be concentrators in level j from the considered nodes and connect them to r. (3) Connect the remaining nodes to the new concentrators. 70 I. G-odor, G. Magyar / Computers & Operations Research 32 (2005) 59–86 We have two options for allocation step within Cluster. In the /rst case, we pick a random node and connect it to the median that provides the least cost; then repeat this until all nodes have been allocated to a median. In the second case, we choose the least-cost connection between a node and a median from all possible node-median pairs. Then we repeat this until all nodes have been allocated. At the relocation step, we try out all nodes in each cluster as a candidate for new median and the one providing the least cost will be chosen as the new median in each cluster, separately. The cost-calculation step is based on the real network costs and it is performed only in the a?ected Download 490.36 Kb. Do'stlaringiz bilan baham: |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling