Ambuj Mahanti

PRA * : massively parallel heuristic search

Abstract In this paper we describe a variant of A* search designed to run on the massively parallel, SIMD Connection Machine (CM-2). The algorithm is designed to run in a limited memory by the use of a retraction technique which allows nodes with poor heuristic values to be removed from the open list until such time as they may need reexpansion, more promising paths having failed. Our algorithm, called PRA* (for Parallel Retraction A*), is designed to maximize use of the Connection Machine′s memory and processors. In addition, the algorithm is guaranteed to return an optimal path when an admissible heuristic is used. Results comparing PRA* to Korf′s IDA* for the fifteen puzzle show significantly fewer node expansions for PRA*. In addition, empirical results show significant parallel speedups, indicative of the algorithm′s design for high processor utilization.

Improving Prediction Accuracy in Trust-Aware Recommender Systems

Trust-aware recommender systems are intelligent technology applications that make use of trust information and user personal data in social networks to provide personalized recommendations. Earlier research in trust-aware systems have shown that the ability of trust-based systems to make accurate predictions coupled with their robustness from shilling attacks make them a better alternative than traditional recommender systems. In this paper we propose an approach for improving accuracy of predictions in trust-aware recommender systems. In our approach, we first reconstruct the trust network. Trust network is reconstructed by removing trust links between users having correlation coefficient below a specified threshold value. For prediction calculation we compare three different approaches based on trust and correlation. We show through experiments on real life Epinions data set that our proposed approach of reconstructing the trust network gives substantially better prediction accuracy than the original approach of using all trust statements in the network.

A SIMD approach to parallel heuristic search

Abstract Serial search algorithms often exhibit exponential run times and may require an exponential amount of storage as well. Thus, the design of parallel search algorithms with limited memory is of obvious interest. This paper presents an efficient SIMD parallel algorithm, called IDPS (for iterative-deepening parallel search). At a broad level IDPS is a parallel version of IDA∗. While generically we have called our algorithm an IDPS, performance of four variants of it has been studied through experiments conducted on the well-known test-bed problem for search algorithms, namely the Fifteen Puzzle. During the experiments, data were gathered under two different static load balancing schemes. Under the first scheme, an unnormalized average efficiency of approximately 3 4 was obtained for 4K, 8K, and 16K processors. Under the second scheme, unnormalized average efficiencies of 0.92 and 0.76, and normalized average efficiencies of 0.70 and 0.63 were obtained for 8K and 16K processors, respectively. We show (as shown previously only for MIMD machines) that for admissible search, high average speedup can be obtained for problems of significant size. We believe that this research will enhance AI problem solving using parallel heuristic search algorithms.

Cyber-risk decision models

Security breaches adversely impact profit margins, market capitalization and brand image of an organization. Global organizations resort to the use of technological devices to reduce the frequency of a security breach. To minimize the impact of financial losses from security breaches, we advocate the use of cyber-insurance products. This paper proposes models to help firms decide on the utility of cyber-insurance products and to what extent they can use them. In this paper, we propose a Copula-aided Bayesian Belief Network (CBBN) for cyber-vulnerability assessment (C-VA), and expected loss computation. Taking these as an input and using the concepts of collective risk modeling theory, we also compute the premium that a cyber risk insurer can charge to indemnify cyber losses. Further, to assist cyber risk insurers and to effectively design products, we propose a utility based preferential pricing (UBPP) model. UBPP takes into account risk profiles and wealth of the prospective insured firm before proposing the premium. Display Omitted Proposed Cyber risk insurance products to minimize the impact of financial loss of security breach.Cyber risk insurance products complement security technology.Our proposed Copula aided Bayesian Belief networks model helps to asses cyber risk.Collective risk & Utility Theory used to computes premium for Cyber risk insurance products.Cyber risks mode for to decide to opt for cyber insurance or not for organizations.

PRA*: a memory-limited heuristic search procedure for the Connection Machine

A variant of A* search designed to run on the massively parallel SIMD (single-instruction-stream, multiple-data-steam) Connection Machine is described. The algorithm is designed to run in a limited memory; a retraction technique allows nodes with poor heuristic values to be removed from the open list until such time as they may need reexpansion if more promising paths fail. The algorithm, called PRA* (for parallel retraction A*), takes maximum advantage of the SIMD design of the Connection Machine and is guaranteed to return an optimal path when an admissible heuristic is used. Results comparing PRA* to R. Korfs IDA* (see Artif. Intell. J., vol.27, 1985) for the 15 puzzle show significantly fewer node expansions for PRA*.<<ETX>>

e-Risk Management with Insurance: A Framework Using Copula Aided Bayesian Belief Networks

e-business organizations are heavily dependent on distributed 24X7 robust information computing systems, for their daily operations. To secure distributed online transactions, they spend millions of dollars on firewalls, anti-virus, intrusion detection systems, digital signature and encryption. Nonetheless, a new virus or a clever hacker can easily compromise these deterrents, resulting in losses to the tune of millions of dollars annually. To cope up with the problem, in this work we propose to further enhance their security management by investing in e-risk insurance products as a viable alternative to reduce these individual financial losses. We develop a framework, based on copula aided Bayesian Belief Network (BBN) model, to quantify the risk associated with online business transactions, arising out of a security breach, and thereby help in designing e-insurance products. We have simulated marginal data for each BBN nodes. The Copula model helps in arriving at the joint probability distributions from these marginal data. From the joint distribution data, we arrive at the conditional distribution tables for each node. This is input to the Bayesian Belief Network model. The output is frequency of occurrence of an e-risk event. Frequency of loss multiplied with the expected loss amount, provides the risk premium to be charged by insurance companies.

A technique to support dynamic pricing strategy for differentiated cellular mobile services

With the advent of a multi-operator regime in the sector of cellular mobile services, users became sensitive to the quality of service (QoS), and, as a result service providers started offering differentiated services. Each operator shall ensure the promised QoS to every subscriber even during busy hours of a day when the network operates at the maximum load. During off-peak hours the load on the network decreases. As a result, the network designed to meet the QoS requirement during busy hours of a day leaves a large amount idle resource during off-peak hours of the day. In this scenario, a service provider faces the problem of maximizing revenue while satisfying resource and QoS constraints. A dynamic differentiated pricing strategy (DDPS) for the cellular mobile service is a solution to the above problem. In this paper, we have proposed a DDPS which, to the best of our knowledge, is the first of its kind in the literature. We have compared the performance of our proposed solution with the static differentiated pricing strategies (SDPSs) with respect to revenue earned and average network resource utilization. The experimental results show that our proposed solution provides a substantial improvement over the SDPSs

Heuristic search techniques for cell to switch assignment in location area planning for cellular networks

The issue of grouping cells into location areas (LAs), where each LA is serviced by a switch, plays an important role in the planning of cellular networks. It is a combinatorial optimization problem that is known to be NP-hard. This paper proposes a total cost of operation (TCO) minimizing state space search formulation of the problem and a heuristic for assigning cell to switches. The TCO includes both recurring hand-off cost and amortized fixed cost. The proposed heuristic is used with block depth first search (BDFS), and also with iterative deepening A (IDA) to solve the problem. Detailed experiments show that the BDFS has a better performance than IDA with respect to execution time while finding an optimal solution. BDFS also outperforms other existing techniques that are based on meta-heuristics, namely, simulated annealing (SA), genetic algorithm (GA), tabu search (TS), and H-I, in terms of solution quality when they are constrained to run for a user specified time.

Strategies for Effective Shilling Attacks against Recommender Systems

One area of research which has recently gained importance is the security of recommender systems. Malicious users may influence the recommender system by inserting biased data into the system. Such attacks may lead to erosion of user trust in the objectivity and accuracy of the system. In this paper, we propose a new approach for creating attack strategies. Our paper explores the importance of target item and filler items in mounting effective shilling attacks. Unlike previous approaches, we propose strategies built specifically for user based and item based collaborative filtering systems. Our attack strategies are based on intelligent selection of filler items. Filler items are selected on the basis of the target item rating distribution. We show through experiments that our strategies are effective against both user based and item based collaborative filtering systems. Our approach is shown to provide substantial improvement in attack effectiveness over existing attack models.

An mHealth Recommender for Smoking Cessation Using Case Based Reasoning

The possible cause of many life-threatening diseases such as lung cancer and cardiac myopathy lies in the addiction to tobacco. Apart from being responsible for premature deaths, early diseases and reduced immunity, smoking also inhibits physiological and psychological problems in children born from addicted mothers. Though many developed countries have consciously made efforts to curb smoking, for the developing countries, the trend is still on the rise. The promising factor is that there are many people who are willing to quit smoking and are resorting to technology and electronic media for the same. In this paper, we have proposed a unique smoking intervention plan with the help of mobile phones that uses a Case Based Recommender system. Our model resorts to generating finely customized motivational messages depending on the patient profile and delivery of the same via mobile phones.