Dimitris Souravlias

Particle swarm optimization with neighborhood-based budget allocation

The standard particle swarm optimization (PSO) algorithm allocates the total available budget of function evaluations equally and concurrently among the particles of the swarm. In the present work, we propose a new variant of PSO where each particle is dynamically assigned different computational budget based on the quality of its neighborhood. The main goal is to favor particles with high-quality neighborhoods by asynchronously providing them with more function evaluations than the rest. For this purpose, we define quality criteria to assess a neighborhood with respect to the information it possesses in terms of solutions’ quality and diversity. Established stochastic techniques are employed for the final selection among the particles. Different variants are proposed by combining various quality criteria in a single- or multi-objective manner. The proposed approach is assessed on widely used test suites as well as on a set of real-world problems. Experimental evidence reveals the efficiency of the proposed approach and its competitiveness against other PSO-based variants as well as different established algorithms.

One is enough: distributed filtering for duplicate elimination

The growth of online services has created the need for duplicate elimination in high-volume streams of events. The sheer volume of data in applications such as pay-per-click clickstream processing, RSS feed syndication and notification services in social sites such Twitter and Facebook makes traditional centralized solutions hard to scale. In this paper, we propose an approach based on distributed filtering. To this end, we introduce a suite of distributed Bloom filters that exploit different ways of partitioning the event space. To address the continuous nature of event delivery, the filters are extended to support sliding window semantics. Moreover, we examine locality-related tradeoffs and propose a tree-based architecture to allow for duplicate elimination across geographic locations. We cast the design space and present experimental results that demonstrate the pros and cons of our various solutions in different settings.

Circulant weighing matrices: a demanding challenge for parallel optimization metaheuristics

Circulant weighing matrices constitute a special type of combinatorial matrices that have attracted scientific interest for many years. The existence and determination of specific classes of circulant weighing matrices remains an active research area that involves both theoretical algebraic techniques as well as high-performance computational optimization approaches. The present work aims at investigating the potential of four established parallel metaheuristics as well as a special Algorithm Portfolio approach, on solving such problems. For this purpose, the algorithms are applied on a hard circulant weighing matrix existence problem. The obtained results are promising, offering insightful conclusions.

Parallel Algorithm Portfolio with Market Trading-Based Time Allocation

We propose a parallel portfolio of metaheuristic algorithms that adopts a market trading-based time allocation mechanism. This mechanism dynamically allocates the total available execution time of the portfolio by favoring better-performing algorithms. The proposed approach is assessed on a significant Operations Research problem, namely the single-item lot sizing problem with returns and remanufacturing. Experimental evidence suggests that our approach is highly competitive with standard metaheuristics and specialized state-of-the-art algorithms.

Particle swarm optimization with budget allocation through neighborhood ranking

Standard Particle Swarm Optimization (PSO) allocates the total available computational budget, in terms of function evaluations, equally among the particles at each iteration of the algorithm. The present work introduces an alternative, which employs neighborhood ranking for allocating the computational budget to the particles. The proposed PSO variant favors the particles that belong to more promising neighborhoods by providing them with more function evaluations than the rest, based on a stochastic neighborhood selection scheme. Preliminary experimental results on standard test problems reveal that the proposed approach is highly competitive.

On novelty in publish/subscribe delivery

In publish/subscribe systems, users express their interests in specific items of information and get notified when relevant data items are produced. Such systems allow users to stay informed without the need of going through huge amounts of data. However, as the volume of data being created increases, some form of ranking of matched events is needed to avoid overwhelming the users. In this work-in-progress paper, we explore novelty as a ranking criterion. An event is considered novel, if it matches a subscription that has rarely been matched in the past.

Designing bijective S-boxes using Algorithm Portfolios with limited time budgets

Abstract Substitution boxes (S-boxes) are essential parts of symmetric-key cryptosystems. Designing S-boxes with satisfactory nonlinearity and autocorrelation properties is a challenging task for both theoretical algebraic methods and computational optimization algorithms. Algorithm Portfolios (APs) are algorithmic schemes where multiple copies of the same algorithm or different algorithms share the available computational resources, running concurrently or interchangeably on a number of processors. Recently, APs have gained increasing attention due to their remarkable efficiency in multidisciplinary applications. The present work is a preliminary study of parallel APs on the bijective S-boxes design problem. The proposed APs comprise two state-of-the-art heuristic algorithms, namely Simulated Annealing and Tabu Search, and they are parallelized according to the master-slave model without exchange of information among the constituent algorithms. The proposed APs are experimentally assessed on typical problem instances under limited time budgets. Different aspects of their performance is analyzed, suggesting that the considered APs are competitive in terms of solution quality and running time against their constituent algorithms as well as different approaches.

On the Solution of Circulant Weighing Matrices Problems Using Algorithm Portfolios on Multi-core Processors

Research on the existence of specific classes of combinatorial matrices such as the Circulant Weighing Matrices CWMs lies in the core of diverse theoretical and computational efforts. Modern metaheuristics have proved to be valuable tools for solving such problems. Recently, parallel Algorithm Portfolios APs composed of established search algorithms and sophisticated resource allocation procedures offered significant improvements in terms of time efficiency and solution quality. The present work aims at shedding further light on the latent quality of parallel APs on solving CWM problems. For this purpose, new AP configurations are considered along with specialized procedures that can enhance their performance. Experimental evaluation is conducted on a computationally restrictive, yet widely accessible, multi-core processor computational environment. Statistical analysis is used to reveal performance trends and extract useful conclusions.

Smart Traffic Lights: A First Parallel Computing Approach

Optimal traffic light scheduling is a fundamental problem in modern urban areas. It has severe impact on traffic flow management, energy consumption and vehicular emissions, as well as on urban noise. The vast number of traffic lights in modern cities increases the complexity of the scheduling problem and, at the same time, urgently needs for efficient algorithms that optimize the light cycle programs. In this work, we propose a solution for the traffic light scheduling problem by using Differential Evolution, and investigate the benefits of parallelism on this complex problem. For understanding the impact in the city, the popular micro-simulator SUMO is used. We evaluate our approach on close-to-reality problem scenarios consisting of two large urban areas located in the cities of Málaga, Spain, and Paris, France. Our results are promising and encourage further investigation of parallel approaches to enable scalability.

Metaheuristic Optimization for Logistics in Natural Disasters

Logistics in natural disasters or emergencies involve highly complicated optimization problems with diverse characteristics. The contribution of the present paper is twofold. First, it introduces a multi-period model aiming to minimize the shortages of different relief products in a number of affected areas. The relief products are transported via multiple modes of transportation from dispatch centers to these areas, while adhering to traffic restrictions. A test suite of benchmark problems with diverse characteristics is generated from the proposed model and solved to optimality with CPLEX. The test suite is used for benchmarking a number of established metaheuristics. Necessary modifications are introduced in the algorithms, in order to fit the special requirements of the specific problem type. The algorithms’ performance is assessed in terms of solution accuracy with respect to the optimal solutions. Comparisons among the employed metaheuristics offer valuable insight regarding their ability to tackle humanitarian logistics problems.