Sujoy Ghose

Heuristic search in restricted memory (research note)

Abstract This paper presents heuristic search algorithms which work within memory constraints. These algorithms, MA∗ (for ordinary graphs) and MAO∗ (for AND/OR graphs) guarantee admissible solutions within specified memory limitations (above the minimum required). The memory versus node expansions tradeoff is analyzed for the worst case. In the case of ordinary graphs, some experiments using the Fifteen Puzzle problem are carried out under various pruning conditions. These parameterized algorithms are found to encompass a wide class of best first search algorithms.

Reducing reexpansions in iterative-deepening search by controlling cutoff bounds

Abstract It is known that a best-first search algorithm like A∗ [5, 6] requires too much space (which often renders it unusable) and a depth-first search strategy does not guarantee an optimal cost solution. The iterative-deepening algorithm IDA∗ [4] achieves both space and cost optimality for a class of tree searching problems. However, for many other problems, it takes too much of computation time due to excessive reexpansion of nodes. This paper presents a modification of IDA∗ to an admissible iterative depth-first branch and bound algorithm IDA∗_CR for trees which overcomes this drawback of IDA∗ and operates much faster using the same amount of storage. Algorithm IDA∗_CRA, a bounded suboptimal cost variation of IDA∗_CR is also presented in order to reduce the execution time still further. Results with the 0/1 Knapsack Problem, Traveling Salesman Problem, and the Flow Shop Scheduling Problem are shown.

GABIND: a GA approach to allocation and binding for the high-level synthesis of data paths

We present here a technique for allocation and binding for data path synthesis (DPS) using a Genetic Algorithm (GA) approach. This GA uses an unconventional crossover mechanism relying on a force directed data path binding completion algorithm. The data path is synthesized using some supplied design parameters. A bus-based interconnection scheme, use of multi-port memories, and provision for multicycling and pipelining are the main features of this system. The method presented here has been applied to standard benchmark examples and the results obtained are promising.

A stochastic model of e-customer behavior

Abstract Web usage mining techniques are increasingly used today to understand e-customers’ within-site behavior. We propose a data mining model that considers e-customers’ activities as a discrete-time semi-Markov process and explains their behavior. An algorithm is proposed to compute transition probability matrix and holding time mass functions from the site navigation data. Finally, the model is used to explain customer behavior in an example site. A software agent, implemented in the site, collects and stores navigation data in the required form and thus helps to avoid data preprocessing. The model results helped to improve the site design and judge its performance.

Point matching using asymmetric neural networks

Abstract Matching control points is an important step in many pattern recognition applications. The matching problem is formulated under translation and rotation as a 0–1 integer programming problem and an artificial neural network is proposed for approximately solving it. The solution to the 0–1 integer programming problem is obtained as the high gain limit point of the continuous network. The network is capable of handling distortion and noise and can use both interpoint distance information and feature properties associated with the points. The results obtained by the network compare favourably with that of the relaxation method of Ton and Jain (IEEE Trans. Geosci. Remote Sensing 27, 642–651 (1989)).

Complexity of fragmentable object bin packing and an application

Abstract We examine in this paper a variant of the bin packing problem, where it is permissible to fragment the objects while packing them into bins of fixed capacity. We call this the Fragmentable Object Bin Packing problem (FOBP). Fragmentation is associated with a cost, leading to the consumption of additional bin capacity. We show that the problem and its absolute approximation are both NP-complete. This is an interesting problem because if the cost of fragmentation is nullified then the problem can be easily solved optimally. If fragmentation is not permitted, then we get the usual bin packing problem. The application comes from a problem in data path synthesis where it is some times necessary to schedule data transfers, subject to restrictions arising from the underlying hardware. We show that FOBP reduces to a simplified version of this problem, thereby proving that it is also a similar hard problem.

Improving greedy algorithms by lookahead-search

Abstract This paper shows that repeated application of a greedy approximation algorithm on some suitably selected subproblems of a problem often leads to a solution which is better than the solution produced by the greedy algorithm applied to the original problem. The lookahead search technique, a polynomial time algorithm introduced here, describes how a greedy algorithm can be utilized in a search process in order to improve the quality of the solution. For the 0/1 knapsack problem and the problem of scheduling independent tasks the lookahead technique is shown to guarantee ϵ-bounded solutions. For the problem of scheduling independent tasks, it has been established that even the simplified version of the lookahead technique provides a bound which is strictly better than the greedy algorithm used in lookahead search. Experimental results are shown for 0/1 knapsack problem, bin packing, Euclidean TSP, and the problem of scheduling independent tasks.

Multihop Virtual Topology Design in WDM Optical Networks for Self-Similar Traffic

In this paper, we consider the problem of designing virtual topologies for multihop optical WDM networks when the traffic is self-similar in nature. Studies over the last few years suggest that the network traffic is bursty and can be much better modeled using self similar process instead of Poisson process. We examine buffer sizes of a network and observe that, even with reasonably low buffer overflow probability, the maximum buffer size requirement for self-similar traffic can be very large. Therefore, a self-similar traffic model has an impact on the queuing delay which is usually much higher than that obtained with the Poisson model. We investigate the problem of constructing the virtual topology with these two types of traffic and solve it with two algorithmic approaches: Greedy (Heuristic) algorithm and Evolutionary algorithm. While the greedy algorithm performs a least-cost search on the total delay along paths for routing traffic in a multihop fashion, the evolutionary algorithm uses genetic methods to optimize the average delay in a network. We analyze and compare our proposed algorithms with an existing algorithm via different performance parameters. Interestingly, with both the proposed algorithms the difference in the queuing delays, caused by self-similar and Poisson traffic, results in different multihop virtual topologies.

Solving constraint optimization problems from CLP-style specifications using heuristic search techniques

Presents a framework for efficiently solving logic formulations of combinatorial optimization problems using heuristic search techniques. In order to integrate cost, lower-bound and upper-bound specifications with conventional logic programming languages, we augment a constraint logic programming (CLP) language with embedded constructs for specifying the cost function and with a few higher-order predicates for specifying the lower and upper bound functions. We illustrate how this simple extension vastly enhances the ease with which optimization problems involving combinations of Min and Max can be specified in the extended language CLP* and we show that CSLDNF (Constraint SLD resolution with Negation as Failure) resolution schemes are not efficient for solving optimization problems specified in this language. Therefore, we describe how any problem specified using CLP* can be converted into an implicit AND/OR graph, and present an algorithm called GenSolve which can branch-and-bound using upper and lower bound estimates, thus exploiting the full pruning power of heuristic search techniques. A technical analysis of GenSolve is provided. We also provide experimental results comparing various control strategies for solving CLP* programs.

Online customized index synthesis in commercial Web sites

An algorithm that can generate and display helpful links while users navigate a site can increase a Web sites usability and help Web designers and the user achieve their goals.