Arquitectura de supercomputación en la nube para la resolución de problemas de optimización combinatoria

La tesis está focalizada en la resolución de problemas de optimización combinatoria, haciendo uso de las opciones tecnológicas actuales que ofrecen las tecnologías de la información y las comunicaciones, y la investigación operativa. Los problemas de optimización combinatoria se resuelven en general mediante programación lineal y metaheurísticas. La aplicación de las técnicas de resolución de los problemas de optimización combinatoria requiere de una elevada carga computacional, y los algoritmos deben diseñarse, por un lado pensando en la efectividad para encontrar buenas soluciones del problema, y por otro lado, pensando en un uso adecuado de los recursos informáticos disponibles. La programación lineal y las metaheurísticas son técnicas de resolución genéricas, que se pueden aplicar a diferentes problemas, partiendo de una base común que se particulariza para cada problema concreto. En el campo del desarrollo de software, los frameworks cumplen esa función de comenzar un proyecto con el trabajo general ya disponible, con la opción de cambiar o extender ese comportamiento base o genérico, para construir el sistema concreto, lo que permite reducir el tiempo de desarrollo, y amplía las posibilidades de éxito del proyecto. En esta tesis se han desarrollado dos frameworks de desarrollo. El framework ILP permite modelar y resolver problemas de programación lineal, de forma independiente al software de resolución de programación lineal que se utilice. El framework LME permite resolver problemas de optimización combinatoria mediante metaheurísticas. Tradicionalmente, las aplicaciones de resolución de problemas de optimización combinatoria son aplicaciones de escritorio que permiten gestionar toda la información de entrada del problema y resuelven el problema en local, con los recursos hardware disponibles. Recientemente ha aparecido un nuevo paradigma de despliegue y uso de aplicaciones que permite compartir recursos informáticos especializados por Internet. Esta nueva forma de uso de recursos informáticos es la computación en la nube, que presenta el modelo de software como servicio (SaaS). En esta tesis se ha construido una plataforma SaaS, para la resolución de problemas de optimización combinatoria, que se despliega sobre arquitecturas compuestas por procesadores multi-núcleo y tarjetas gráficas, y dispone de algoritmos de resolución basados en frameworks de programación lineal y metaheurísticas. Toda la infraestructura es independiente del problema de optimización combinatoria a resolver, y se han desarrollado tres problemas que están totalmente integrados en la plataforma SaaS. Estos problemas se han seleccionado por su importancia práctica. Uno de los problemas tratados en la tesis, es el problema de rutas de vehículos (VRP), que consiste en calcular las rutas de menor coste de una flota de vehículos, que reparte mercancías a todos los clientes. Se ha partido de la versión más clásica del problema y se han hecho estudios en dos direcciones. Por un lado se ha cuantificado el aumento en la velocidad de ejecución de la resolución del problema en tarjetas gráficas. Por otro lado, se ha estudiado el impacto en la velocidad de ejecución y en la calidad de soluciones, en la resolución por la metaheurística de colonias de hormigas (ACO), cuando se introduce la programación lineal para optimizar las rutas individuales de cada vehículo. Este problema se ha desarrollado con los frameworks ILP y LME, y está disponible en la plataforma SaaS. Otro de los problemas tratados en la tesis, es el problema de asignación de flotas (FAP), que consiste en crear las rutas de menor coste para la flota de vehículos de una empresa de transporte de viajeros. Se ha definido un nuevo modelo de problema, que engloba características de problemas presentados en la literatura, y añade nuevas características, lo que permite modelar los requerimientos de las empresas de transporte de viajeros actuales. Este nuevo modelo resuelve de forma integrada el problema de definir los horarios de los trayectos, el problema de asignación del tipo de vehículo, y el problema de crear las rotaciones de los vehículos. Se ha creado un modelo de programación lineal para el problema, y se ha resuelto por programación lineal y por colonias de hormigas (ACO). Este problema se ha desarrollado con los frameworks ILP y LME, y está disponible en la plataforma SaaS. El último problema tratado en la tesis es el problema de planificación táctica de personal (TWFP), que consiste en definir la configuración de una plantilla de trabajadores de menor coste, para cubrir una demanda de carga de trabajo variable. Se ha definido un modelo de problema muy flexible en la definición de contratos, que permite el uso del modelo en diversos sectores productivos. Se ha definido un modelo matemático de programación lineal para representar el problema. Se han definido una serie de casos de uso, que muestran la versatilidad del modelo de problema, y permiten simular el proceso de toma de decisiones de la configuración de una plantilla de trabajadores, cuantificando económicamente cada decisión que se toma. Este problema se ha desarrollado con el framework ILP, y está disponible en la plataforma SaaS. ABSTRACT The thesis is focused on solving combinatorial optimization problems, using current technology options offered by information technology and communications, and operations research. Combinatorial optimization problems are solved in general by linear programming and metaheuristics. The application of these techniques for solving combinatorial optimization problems requires a high computational load, and algorithms are designed, on the one hand thinking to find good solutions to the problem, and on the other hand, thinking about proper use of the available computing resources. Linear programming and metaheuristic are generic resolution techniques, which can be applied to different problems, beginning with a common base that is particularized for each specific problem. In the field of software development, frameworks fulfill this function that allows you to start a project with the overall work already available, with the option to change or extend the behavior or generic basis, to build the concrete system, thus reducing the time development, and expanding the possibilities of success of the project. In this thesis, two development frameworks have been designed and developed. The ILP framework allows to modeling and solving linear programming problems, regardless of the linear programming solver used. The LME framework is designed for solving combinatorial optimization problems using metaheuristics. Traditionally, applications for solving combinatorial optimization problems are desktop applications that allow the user to manage all the information input of the problem and solve the problem locally, using the available hardware resources. Recently, a new deployment paradigm has appeared, that lets to share hardware and software resources by the Internet. This new use of computer resources is cloud computing, which presents the model of software as a service (SaaS). In this thesis, a SaaS platform has been built for solving combinatorial optimization problems, which is deployed on architectures, composed of multi-core processors and graphics cards, and has algorithms based on metaheuristics and linear programming frameworks. The SaaS infrastructure is independent of the combinatorial optimization problem to solve, and three problems are fully integrated into the SaaS platform. These problems have been selected for their practical importance. One of the problems discussed in the thesis, is the vehicle routing problem (VRP), which goal is to calculate the least cost of a fleet of vehicles, which distributes goods to all customers. The VRP has been studied in two directions. On one hand, it has been quantified the increase in execution speed when the problem is solved on graphics cards. On the other hand, it has been studied the impact on execution speed and quality of solutions, when the problem is solved by ant colony optimization (ACO) metaheuristic, and linear programming is introduced to optimize the individual routes of each vehicle. This problem has been developed with the ILP and LME frameworks, and is available in the SaaS platform. Another problem addressed in the thesis, is the fleet assignment problem (FAP), which goal is to create lower cost routes for a fleet of a passenger transport company. It has been defined a new model of problem, which includes features of problems presented in the literature, and adds new features, allowing modeling the business requirements of today's transport companies. This new integrated model solves the problem of defining the flights timetable, the problem of assigning the type of vehicle, and the problem of creating aircraft rotations. The problem has been solved by linear programming and ACO. This problem has been developed with the ILP and LME frameworks, and is available in the SaaS platform. The last problem discussed in the thesis is the tactical planning staff problem (TWFP), which is to define the staff of lower cost, to cover a given work load. It has been defined a very rich problem model in the definition of contracts, allowing the use of the model in various productive sectors. It has been defined a linear programming mathematical model to represent the problem. Some use cases has been defined, to show the versatility of the model problem, and to simulate the decision making process of setting up a staff, economically quantifying every decision that is made. This problem has been developed with the ILP framework, and is available in the SaaS platform.

Global behavior of the Douglas–Rachford method for a nonconvex feasibility problem

In recent times the Douglas–Rachford algorithm has been observed empirically to solve a variety of nonconvex feasibility problems including those of a combinatorial nature. For many of these problems current theory is not sufficient to explain this observed success and is mainly concerned with questions of local convergence. In this paper we analyze global behavior of the method for finding a point in the intersection of a half-space and a potentially non-convex set which is assumed to satisfy a well-quasi-ordering property or a property weaker than compactness. In particular, the special case in which the second set is finite is covered by our framework and provides a prototypical setting for combinatorial optimization problems.

Optimization of ATM Telecommunication Networks

ATM network optimization problems defined as combinatorial optimization problems are considered. Several approximate algorithms for solving such problems are developed. Results of their comparison by experiments on a set of problems with random input data are presented.

Differential Balanced Trees and (0,1) Matrices

* The research was supported by INTAS 00-397 and 00-626 Projects.

Approximation Algorithms for Network Design and Orienteering

This thesis presents approximation algorithms for some NP-Hard combinatorial optimization problems on graphs and networks; in particular, we study problems related to Network Design. Under the widely-believed complexity-theoretic assumption that P is not equal to NP, there are no efficient (i.e., polynomial-time) algorithms that solve these problems exactly. Hence, if one desires efficient algorithms for such problems, it is necessary to consider approximate solutions: An approximation algorithm for an NP-Hard problem is a polynomial time algorithm which, for any instance of the problem, finds a solution whose value is guaranteed to be within a multiplicative factor of the value of an optimal solution to that instance. We attempt to design algorithms for which this factor, referred to as the approximation ratio of the algorithm, is as small as possible. The field of Network Design comprises a large class of problems that deal with constructing networks of low cost and/or high capacity, routing data through existing networks, and many related issues. In this thesis, we focus chiefly on designing fault-tolerant networks. Two vertices u,v in a network are said to be k-edge-connected if deleting any set of k − 1 edges leaves u and v connected; similarly, they are k-vertex connected if deleting any set of k − 1 other vertices or edges leaves u and v connected. We focus on building networks that are highly connected, meaning that even if a small number of edges and nodes fail, the remaining nodes will still be able to communicate. A brief description of some of our results is given below. We study the problem of building 2-vertex-connected networks that are large and have low cost. Given an n-node graph with costs on its edges and any integer k, we give an O(log n log k) approximation for the problem of finding a minimum-cost 2-vertex-connected subgraph containing at least k nodes. We also give an algorithm of similar approximation ratio for maximizing the number of nodes in a 2-vertex-connected subgraph subject to a budget constraint on the total cost of its edges. Our algorithms are based on a pruning process that, given a 2-vertex-connected graph, finds a 2-vertex-connected subgraph of any desired size and of density comparable to the input graph, where the density of a graph is the ratio of its cost to the number of vertices it contains. This pruning algorithm is simple and efficient, and is likely to find additional applications. Recent breakthroughs on vertex-connectivity have made use of algorithms for element-connectivity problems. We develop an algorithm that, given a graph with some vertices marked as terminals, significantly simplifies the graph while preserving the pairwise element-connectivity of all terminals; in fact, the resulting graph is bipartite. We believe that our simplification/reduction algorithm will be a useful tool in many settings. We illustrate its applicability by giving algorithms to find many trees that each span a given terminal set, while being disjoint on edges and non-terminal vertices; such problems have applications in VLSI design and other areas. We also use this reduction algorithm to analyze simple algorithms for single-sink network design problems with high vertex-connectivity requirements; we give an O(k log n)-approximation for the problem of k-connecting a given set of terminals to a common sink. We study similar problems in which different types of links, of varying capacities and costs, can be used to connect nodes; assuming there are economies of scale, we give algorithms to construct low-cost networks with sufficient capacity or bandwidth to simultaneously support flow from each terminal to the common sink along many vertex-disjoint paths. We further investigate capacitated network design, where edges may have arbitrary costs and capacities. Given a connectivity requirement R_uv for each pair of vertices u,v, the goal is to find a low-cost network which, for each uv, can support a flow of R_uv units of traffic between u and v. We study several special cases of this problem, giving both algorithmic and hardness results. In addition to Network Design, we consider certain Traveling Salesperson-like problems, where the goal is to find short walks that visit many distinct vertices. We give a (2 + epsilon)-approximation for Orienteering in undirected graphs, achieving the best known approximation ratio, and the first approximation algorithm for Orienteering in directed graphs. We also give improved algorithms for Orienteering with time windows, in which vertices must be visited between specified release times and deadlines, and other related problems. These problems are motivated by applications in the fields of vehicle routing, delivery and transportation of goods, and robot path planning.

Data-Aware Scheduling in Datacenters

Datacenters have emerged as the dominant form of computing infrastructure over the last two decades. The tremendous increase in the requirements of data analysis has led to a proportional increase in power consumption and datacenters are now one of the fastest growing electricity consumers in the United States. Another rising concern is the loss of throughput due to network congestion. Scheduling models that do not explicitly account for data placement may lead to a transfer of large amounts of data over the network causing unacceptable delays. In this dissertation, we study different scheduling models that are inspired by the dual objectives of minimizing energy costs and network congestion in a datacenter. As datacenters are equipped to handle peak workloads, the average server utilization in most datacenters is very low. As a result, one can achieve huge energy savings by selectively shutting down machines when demand is low. In this dissertation, we introduce the network-aware machine activation problem to find a schedule that simultaneously minimizes the number of machines necessary and the congestion incurred in the network. Our model significantly generalizes well-studied combinatorial optimization problems such as hard-capacitated hypergraph covering and is thus strongly NP-hard. As a result, we focus on finding good approximation algorithms. Data-parallel computation frameworks such as MapReduce have popularized the design of applications that require a large amount of communication between different machines. Efficient scheduling of these communication demands is essential to guarantee efficient execution of the different applications. In the second part of the thesis, we study the approximability of the co-flow scheduling problem that has been recently introduced to capture these application-level demands. Finally, we also study the question, "In what order should one process jobs?'' Often, precedence constraints specify a partial order over the set of jobs and the objective is to find suitable schedules that satisfy the partial order. However, in the presence of hard deadline constraints, it may be impossible to find a schedule that satisfies all precedence constraints. In this thesis we formalize different variants of job scheduling with soft precedence constraints and conduct the first systematic study of these problems.

An experimental study of adaptive control for evolutionary algorithms

International audience

Boosting the performance of metaheuristics for the MinLA problem using a more discriminating evaluation function

International audience

Mathematical Programming Models for Influence Maximization on Social Networks

In this dissertation, we apply mathematical programming techniques (i.e., integer programming and polyhedral combinatorics) to develop exact approaches for influence maximization on social networks. We study four combinatorial optimization problems that deal with maximizing influence at minimum cost over a social network. To our knowl- edge, all previous work to date involving influence maximization problems has focused on heuristics and approximation. We start with the following viral marketing problem that has attracted a significant amount of interest from the computer science literature. Given a social network, find a target set of customers to seed with a product. Then, a cascade will be caused by these initial adopters and other people start to adopt this product due to the influence they re- ceive from earlier adopters. The idea is to find the minimum cost that results in the entire network adopting the product. We first study a problem called the Weighted Target Set Selection (WTSS) Prob- lem. In the WTSS problem, the diffusion can take place over as many time periods as needed and a free product is given out to the individuals in the target set. Restricting the number of time periods that the diffusion takes place over to be one, we obtain a problem called the Positive Influence Dominating Set (PIDS) problem. Next, incorporating partial incentives, we consider a problem called the Least Cost Influence Problem (LCIP). The fourth problem studied is the One Time Period Least Cost Influence Problem (1TPLCIP) which is identical to the LCIP except that we restrict the number of time periods that the diffusion takes place over to be one. We apply a common research paradigm to each of these four problems. First, we work on special graphs: trees and cycles. Based on the insights we obtain from special graphs, we develop efficient methods for general graphs. On trees, first, we propose a polynomial time algorithm. More importantly, we present a tight and compact extended formulation. We also project the extended formulation onto the space of the natural vari- ables that gives the polytope on trees. Next, building upon the result for trees---we derive the polytope on cycles for the WTSS problem; as well as a polynomial time algorithm on cycles. This leads to our contribution on general graphs. For the WTSS problem and the LCIP, using the observation that the influence propagation network must be a directed acyclic graph (DAG), the strong formulation for trees can be embedded into a formulation on general graphs. We use this to design and implement a branch-and-cut approach for the WTSS problem and the LCIP. In our computational study, we are able to obtain high quality solutions for random graph instances with up to 10,000 nodes and 20,000 edges (40,000 arcs) within a reasonable amount of time.

Methods for integrating machine learning and constrained optimization

In the framework of industrial problems, the application of Constrained Optimization is known to have overall very good modeling capability and performance and stands as one of the most powerful, explored, and exploited tool to address prescriptive tasks. The number of applications is huge, ranging from logistics to transportation, packing, production, telecommunication, scheduling, and much more. The main reason behind this success is to be found in the remarkable effort put in the last decades by the OR community to develop realistic models and devise exact or approximate methods to solve the largest variety of constrained or combinatorial optimization problems, together with the spread of computational power and easily accessible OR software and resources. On the other hand, the technological advancements lead to a data wealth never seen before and increasingly push towards methods able to extract useful knowledge from them; among the data-driven methods, Machine Learning techniques appear to be one of the most promising, thanks to its successes in domains like Image Recognition, Natural Language Processes and playing games, but also the amount of research involved. The purpose of the present research is to study how Machine Learning and Constrained Optimization can be used together to achieve systems able to leverage the strengths of both methods: this would open the way to exploiting decades of research on resolution techniques for COPs and constructing models able to adapt and learn from available data. In the first part of this work, we survey the existing techniques and classify them according to the type, method, or scope of the integration; subsequently, we introduce a novel and general algorithm devised to inject knowledge into learning models through constraints, Moving Target. In the last part of the thesis, two applications stemming from real-world projects and done in collaboration with Optit will be presented.

Algorithmic Solutions for Combinatorial Problems in Resource Management of Manufacturing Environments

This thesis studies the use of heuristic algorithms in a number of combinatorial problems that occur in various resource constrained environments. Such problems occur, for example, in manufacturing, where a restricted number of resources (tools, machines, feeder slots) are needed to perform some operations. Many of these problems turn out to be computationally intractable, and heuristic algorithms are used to provide efficient, yet sub-optimal solutions. The main goal of the present study is to build upon existing methods to create new heuristics that provide improved solutions for some of these problems. All of these problems occur in practice, and one of the motivations of our study was the request for improvements from industrial sources. We approach three different resource constrained problems. The first is the tool switching and loading problem, and occurs especially in the assembly of printed circuit boards. This problem has to be solved when an efficient, yet small primary storage is used to access resources (tools) from a less efficient (but unlimited) secondary storage area. We study various forms of the problem and provide improved heuristics for its solution. Second, the nozzle assignment problem is concerned with selecting a suitable set of vacuum nozzles for the arms of a robotic assembly machine. It turns out that this is a specialized formulation of the MINMAX resource allocation formulation of the apportionment problem and it can be solved efficiently and optimally. We construct an exact algorithm specialized for the nozzle selection and provide a proof of its optimality. Third, the problem of feeder assignment and component tape construction occurs when electronic components are inserted and certain component types cause tape movement delays that can significantly impact the efficiency of printed circuit board assembly. Here, careful selection of component slots in the feeder improves the tape movement speed. We provide a formal proof that this problem is of the same complexity as the turnpike problem (a well studied geometric optimization problem), and provide a heuristic algorithm for this problem.

Approximating a class of combinatorial problems with rational objective function

In the late seventies, Megiddo proposed a way to use an algorithm for the problem of minimizing a linear function a(0) + a(1)x(1) + ... + a(n)x(n) subject to certain constraints to solve the problem of minimizing a rational function of the form (a(0) + a(1)x(1) + ... + a(n)x(n))/(b(0) + b(1)x(1) + ... + b(n)x(n)) subject to the same set of constraints, assuming that the denominator is always positive. Using a rather strong assumption, Hashizume et al. extended Megiddo`s result to include approximation algorithms. Their assumption essentially asks for the existence of good approximation algorithms for optimization problems with possibly negative coefficients in the (linear) objective function, which is rather unusual for most combinatorial problems. In this paper, we present an alternative extension of Megiddo`s result for approximations that avoids this issue and applies to a large class of optimization problems. Specifically, we show that, if there is an alpha-approximation for the problem of minimizing a nonnegative linear function subject to constraints satisfying a certain increasing property then there is an alpha-approximation (1 1/alpha-approximation) for the problem of minimizing (maximizing) a nonnegative rational function subject to the same constraints. Our framework applies to covering problems and network design problems, among others.

Vector Combinatorial Problems in a Space of Combinations with Linear Fractional Functions of Criteria

The paper considers vector discrete optimization problem with linear fractional functions of criteria on a feasible set that has combinatorial properties of combinations. Structural properties of a feasible solution domain and of Pareto–optimal (efficient), weakly efficient, strictly efficient solution sets are examined. A relation between vector optimization problems on a combinatorial set of combinations and on a continuous feasible set is determined. One possible approach is proposed in order to solve a multicriteria combinatorial problem with linear- fractional functions of criteria on a set of combinations.

Constructing a Canonical form of a Matrix in Several Problems about Combinatorial Designs

Partially supported by the Bulgarian Science Fund contract with TU Varna, No 487.

Representing Equivalence Problems for Combinatorial Objects

Methods for representing equivalence problems of various combinatorial objects as graphs or binary matrices are considered. Such representations can be used for isomorphism testing in classification or generation algorithms. Often it is easier to consider a graph or a binary matrix isomorphism problem than to implement heavy algorithms depending especially on particular combinatorial objects. Moreover, there already exist well tested algorithms for the graph isomorphism problem (nauty) and the binary matrix isomorphism problem as well (Q-Extension). ACM Computing Classification System (1998): F.2.1, G.4.