Link enhancement using constrained range and reduced candidate set searches: a revision

Ruey-yih Lin, Chyan Yang
1995 Computer Communications  
This paper revises the CRCS algorithm proposed by Yang'. In the constrained range search, we use the cost and performance of candidate links to obtain the exact link range of the optimal solution. Moreover, we use the concept of variable dominance to reduce the number of candidate links without loss of optimality. A comparison of results obtained using Yang's and the revised CRCS algorithm shows that the revised CRCS algorithm is superior to Yang's original version. Keywords: candidate set
more » ... h, constrained range search, variable dominance In a recent paper in this journal, Yang' proposed a heuristic algorithm that uses constrained range and reduced candidate set search (CRCS) to reduce computation time while obtaining a near-optimal solution. The key finding in Yang's paper is that the CRCS algorithm is justified, since the algorithm takes only a non-exponential time to compute and has a high probability of reaching optimality. However, Yang's work raises two problems that need to be investigated. First, the range of the con-strained range search in Yang's paper, which uses the costs of candidate links to obtain the range (or number of links) of the optimum solution, is very wide, so a further 'squeeze' is required. Second, since Yang's method chooses a candidate set through a linear search algorithm, it is possible that optimality will be lost, i.e. the reduced candidate set in Yang's method could rule out a link which is in the optimal solution set. The purpose of this paper is to revise the CRCS algorithm. Two major improvements are proposed: first, by combining information on the cost and performance of candidate links, we obtain the exact number of links in one step without compromising optimality and without requiring further 'squeeze'; second, instead of using a linear search algorithm, we use the concept of variable dominance to reduce the candidate set and ensure that the optimal solution is in the reduced set. Further, we include the dominance of variables in the computation to increase searching efficiency. The rest of this paper is organized as follows. The revised CRCS algorithm is presented in the next section; the following section compares the results obtained using Yang's CRCS algorithm with those obtained using the revised algorithm. Conclusions and the limitations of CRCS algorithms are discussed in the final section. REVISION OF THE CRCS ALGORITHM Two major improvements can be made to Yang's CRCS algorithm. The first is to tighten the feasible space, which is determined by the available budget and the given cost of the links. The second is to use the concept of variable dominance to retain (or rule out) links that are definitely (or definitely not) in the optimal solution set without losing optimality. We shall call the revised CRCS method CRCS'. We explain this search method as follows. New constrained range search First, we describe the method of constraining the range in the CRCS' algorithm. Given a budget B and the cost Ci and performance Pi of each candidate link i, we may squeeze the number of links within a constrained range without compromising optimality, hence reducing the computational cost tremendously. The revised constrained range search algorithm can tighten the number of optimal links exactly as is done in Yang's examples. First, 0140-3664/95/$09.50 0 1995-Elsevier Science B.
doi:10.1016/0140-3664(95)99817-v fatcat:drfnstv5z5brzhyb5sqqwihbji