MIC'2017: 12TH METAHEURISTICS INTERNATIONAL CONFERENCE
PROGRAM FOR FRIDAY, JULY 7TH
Days:
previous day
all days

View: session overviewtalk overview

09:00-10:00 Session F1.1: Metaheuristics and Machine Learning
Location: 20.017
09:00
Learnheuristics: hybridizing metaheuristics with machine learning for optimization with dynamic inputs
SPEAKER: Laura Calvet

ABSTRACT. This paper reviews the existing literature on the combination of metaheuristics with machine learning methods and then introduces the concept of learnheuristics, a novel type of hybrid algorithms. Learnheuristics can be used to solve combinatorial optimization problems with dynamic inputs (COPDIs). In these COPDIs, the problem inputs (elements either located in the objective function or in the constraints set) are not fixed in advance as usual. On the contrary, they might vary in a predictable (non-random) way as the solution is partially built according to some heuristic-based iterative process. For instance, a consumer’s willingness to spend on a specific product might change as the availability of this product decreases and its price rises. Thus, these inputs might take different values depending on the current solution configuration. These variations in the inputs might require from a coordination between the learning mechanism and the metaheuristic algorithm: at each iteration, the learning method updates the inputs model used by the metaheuristic.

09:20
Setting the Number of Simulation Runs in a Simheuristic Algorithm: a statistical-based approach

ABSTRACT. Many real-life systems in logistics and transportation are complex, large-scale, and stochastic in nature. As a consequence, simheuristic approaches that combine metaheuristics with simulation are becoming increasingly popular in the optimization and simulation research. In a simheuristic algorithm, time-consuming simulation runs are typically required in order to obtain accurate results. If the underlying system is complex, the simulation component could jeopardize the computing time of the metaheuristic component. Thus, for each new solution generated by the metaheuristic component, two related questions arise: (i) should the simulation component be applied on that solution? and (ii) if so, how many simulation runs are needed in order to obtain ‘accurate’ results? This paper proposes a methodology, based on the combined use of deterministic runs and confidence intervals, that allows answering both questions. Computational experiments contribute to illustrate our approach and also to analyze how the number or required runs varies as different significance levels are considered.

09:40
Artificial Immunne Systems based Astronomy Scheduling

ABSTRACT. Planning astronomy observations requires intensive work to adapt the observation plans to the changing conditions of observation projects (i.e., weather conditions, source visibility, technical failures). In addition, some schedulers/planners that use machine learning techniques require complex sample data of observation sequences, which are usually not available for most of the astronomy telescopes. In addition, since traditional scheduling methods are unable to self-organize, they usually require effort to optimize system parameters. In order to address these issues, in this work a new method is proposed to schedule astronomy observations projects. The approach uses {\em artificial immune systems} techniques in order to optimize observation plans and available resources according to real-time scientific priorities. Experiments using real and synthetic data on observation proposals and weather information, show the promise of the method when compared with traditional scheduling algorithms.

09:00-10:00 Session F1.2: Metaheuristics in Education
Location: 20.019
09:00
Introducing Metaheuristics in an e-learning Course on Distributed Systems

ABSTRACT. Distributed Systems is a compulsory subject in the Computer Science degree of most universities due to the emergence of new computing paradigms in recent years. One of the main challenges in this field is to assign jobs to the most suitable resources in order to run them in an efficient way. Given the growing popularity of metaheuristics to address this problem, we have decided to include them in the programme of the subject at the Open University of Catalonia, an online university. This work describes the first experiences and shares insights into the design and development of the course.

09:20
Teaching metaheuristics online: experiences at the Open University of Catalonia
SPEAKER: unknown

ABSTRACT. Metaheuristics have become the first-resource methodology for solving real and complex combinatorial optimization problems (COPs), being highly employed by researchers and companies. Accordingly, they have been introduced in study programs of most universities. This work describes experiences of a course in metaheuristics at the Open University of Catalonia (UOC), an online university. After listing the main characteristics of e-learning, we present the UOC and the course, discussing learning issues as the suitability of working on classical COPs of different fields, developing both analytical and computational skills, and reading about recent investigations. Finally, we describe challenges encountered and how they have been addressed, and discuss potential improvements.

09:00-10:00 Session F1.3: Metaheuristics in Port Logistics and Related Problems
Location: 20.021
09:00
A General Framework for Handling Synchronization when Scheduling Automated Guided Vehicles

ABSTRACT. The present paper introduces and solves the synchronization problems associated with the evaluation of solutions of the Internal Delivery Vehicle Synchronization Problem. This optimization problem addresses the management of a fleet of automated guided vehicles to perform the movement of containers around a maritime container terminal. Synchronization between vehicles is required when they seek to retrieve or store containers in the same location at once. A framework aimed at handling the synchronization problems is here proposed. This includes a set of strategies to handle the priority of the vehicles in those scenarios in which synchronization is required. An empirical study is performed to check the performance of these strategies.

09:20
Programación de un sistema portuario implementando algoritmos genéticos con un modelo de simulación

ABSTRACT. Este trabajo aborda el problema de asignación de atraque (Berth Allocation Problem – BAP) en un sistema portuario. Los sistemas portuarios más comunes en la práctica son las terminales de contenedores y las terminales graneleras. Las terminales de contenedores son aquellas en que la unidad de manejo del puerto es el contenedor, estas terminales necesitan recursos específicos para realizar el cargue y descargue como las grúas pórtico. Las terminales graneleras pueden realizar el cargue a través de un dispositivo de cargue llamado shiploader para manejar bienes al granel tales como carbón, hierro, cobre, entre otras. Los problemas de asignación de atraque son problemas NP-duros que se han tratado de forma discreta y continua. Se han generado una variedad de formulaciones considerado restricciones tales como el layout. Al no existir un algoritmo exacto que permita resolver este problema en un tiempo razonable por su alta complejidad computacional, varios autores han implementado diferentes metaheurísticas para tratarlo, entre las cuales se encuentran las siguientes: algoritmos genéticos, búsqueda tabú, optimización por colonias de hormigas, GRASP, recocido simulado, búsqueda de vecindad variable y optimización por enjambres de partículas. Las investigaciones realizadas al problema de asignación de atraque que utilizan técnicas de simulación optimización son muy pocas. Por décadas se ha considerado la simulación como un método de última instancia, solamente utilizada cuando todo lo demás no funciona. Sin embargo, el desarrollo de hardware con mayores capacidades y software especializado de simulación han aumentado el potencial de procesamiento de los computadores permitiendo resolver instancias de gran tamaño en tiempos de computación cortos. Este trabajo, que se encuentra en progreso, analiza el problema de programación en sistemas portuarios al granel para minimizar el demurrage, que es una penalización que se incurre por in-cumplir los laydays; siendo éste el tiempo más temprano cuando el buque es requerido por el fletador y el tiempo límite para iniciar el cargue/descargue. Los buques graneleros deben ser atendidos por cargadores idénticos llamados shiploaders. En caso, que el puerto empiece el cargue después de terminar el laydays, el fletador será penalizado por la cantidad de tiempo que transcurre desde la finalización del laydays hasta el inicio del cargue. En el sistema portuario actual, se ha analizado la programación de seis buques con dos shiploaders. Se han comparado dos situaciones en la secuenciación de los buques de carga granel permitiendo medir el demurrage: la situación actual ordenando los buques de acuerdo a las fechas límites para realizar la carga y una implementación de algoritmos genéticos (GA) con el software de simulación Flexsim 7.0.6. Se encontró por medio de la utilización de un diseño experimental factorial, que el Dual Crossover Point es el operador de cruzamiento que genera mejores soluciones para la programación de buques en este puerto de carga a granel. Sin embargo, la programación de un sistema portuario más realista se realiza con un número mayor de buques y de shiploaders conforme a las necesidades portuarias. Por tal razón, dicha investigación no ha concluido. Se hace necesario seguir con este estudio con el propósito de tener resultados de instancias más realistas. Siguiendo este enfoque, se está aplicando una metodología que permite abordar instancias más realistas de este problema con condiciones estocásticas utilizando un modelo simulado. Adicionalmente, con la ayuda de la simulación se ha podido observar otras medidas de desempeño del sistema como es el caso de la utilización de los recursos, que no son fácilmente observables cuando se utilizan heurísticas de búsqueda para encontrar una solución. La presente investigación desarrolla un BAP discreto, considerando los tiempos de cargue/descargue y los tiempos de alistamiento por cambio de material estocásticos en una terminal granelera con el objetivo de minimizar el dumurrage. Se plantea la utilización de otros algoritmos heurísticos propuestos en la literatura para abordar el BAP, para compararlos con los resultados obtenidos por el algoritmo genético planteado. Se probarán un conjunto de instancias para el BAP discreto y se tabularán los resultados. Finalmente, se realizará un análisis de riesgo que per-mita identificar las soluciones de alta calidad encontradas.

09:40
A Biased Random-Key Genetic Algorithm for the Berth Allocation and Quay Crane Assignment Problem

ABSTRACT. The Berth Allocation and Quay Crane Assignment Problem (BACAP) consists in assigning a number of cranes, and a berthing time and position to each calling vessel within a planning horizon so as to minimize the total assignment cost. In this paper, we address both the BACAP variant consisting of a continuous quay, dynamic arrivals and time-invariant cranes-to-vessels assignments. We propose a metaheuristic approach based on a Biased Random-key Genetic Algorithm with memetic characteristics and several Local Search procedures. The performance of this method, in terms of both time and quality of the solutions obtained, was tested in several computational experiments. The results show that our approach is able to find optimal solutions on instances of up to 40 vessels and good solutions on instances of up to 100 vessels.

09:00-10:00 Session F1.4: Metaheuristics - General
Location: 20.023
09:00
An improved Cuckoo Search Algorithm for the cardinality constraint portfolio optimization problem

ABSTRACT. This paper presents a new metaheuristics based on a hybrid combination of Cuckoo Search (CS) and optimization algorithms for solving the cardinality constraint portfolio optimization (CCPO) problem. The paper presents a brief review on existing literature and highlights related issues. A hybrid metaheuristic is then introduced where CS algorithm explores the cardinality search space using Levy flights to select the desired assets for investment, while an exact optimization method determines the optimal allocation of investments for each of the chosen set of assets. The main contributions include an introduction of a new mapping between the continuous search space and the integer search space to guide the CS selection of assets; improvements to the CS components to obtain comparable solution quality to the previously published results on a set of benchmark data set from the literature. Further research directions are also reported

09:20
Can single-objective metaheuristics efficiently solve multi-objective problems?
SPEAKER: Piotr Matl

ABSTRACT. Real-life problems are often characterized by conflicting optimization objectives. As a result, there has been a growing interest not only in multi-objective models, but also in corresponding solution algorithms, especially multi-objective metaheuristics. Yet in principle, multi-objective problems can be efficiently solved with existing tailored single-objective solvers – this is the central idea behind the well-known epsilon-constraint method. Despite its theoretical properties and conceptual simplicity, the epsilon-constraint framework has been largely ignored in the domain of heuristics and remains associated mostly with exact algorithms. In this paper we dispel these preconceptions and demonstrate that the epsilon-constraint method can be a highly effective way to leverage the existing research on single-objective optimization for solving multi-objective problems.

We propose an improved version of the classical epsilon-constraint method which accounts for the challenges and demands specific to heuristic search. The resulting framework is implemented with an existing state-of-the-art CVRP solver and tested on the VRP with Route Balancing. Based on an in-depth computational study, we demonstrate that our generic epsilon-constraint algorithm significantly outperforms two recent state of-the-art multi-objective metaheuristics with respect to multiple quality metrics.

09:40
A local search-based multi-objective metaheuristic for biclustering on heterogeneous data

ABSTRACT. In this article, we address the problem of biclustering on heterogeneous data, that is, data of various types (binary, numeric, etc.). This problem has not yet been investigated in the biclustering literature. We propose a new method, LS-HBC (Local Search-based Heterogeneous BiClustering), designed to extract biclusters from heterogeneous, large-scale, sparse data matrices. This method is well-suited to provide valuable insight on medical data gathered by hospitals (on patients, stays, acts, diagnoses, prescriptions, etc.), but may be used for other applications. LS-HBC is a multi-objective optimization algorithm for biclustering, which generates a set of possibly overlapping high-quality biclusters. Experiments show that this algorithm generates better solutions than a standard biclustering method on small-size numeric data, but suffers from a lack of diversity amongst the archives of solutions created. Experiments on real-life medical data sets further assert its scalability and efficiency.

09:00-10:00 Session F1.5: Metaheuristics General (MAEB)
Location: 20.025
09:00
Análisis de algoritmos de evolución gramatical en problemas de regresión simbólica

ABSTRACT. La evolución gramatical o gramáticas evolutivas (GE, Grammatical Evolution en inglés) son una rama de la programación genética que surgió, hace ahora 20 años, como una forma de evolucionar programas utilizando gramáticas libres de contexto (context free grammars) para generar el código. El objetivo era también reducir el tamaño de los programas y reducir, con respecto a otras aproximaciones, las proporciones de código no efectivo generado en la decodificación. El uso de las gramáticas permite dirigir la búsqueda aportando conocimiento. Por contra, un mal diseño también puede acarrear problemas para alcanzar ciertas partes del espacio de búsqueda. Una de las aplicaciones más fructíferas de las gramáticas evolutivas es la resolución de problemas de regresión simbólica. En este artículo se realiza un análisis práctico de los efectos en el proceso evolutivo de los distintos factores implicados en una gramática evolutiva para un problema de regresión simbólica de 10 variables. El artículo se centrará en el uso de distintas gramáticas, pero también analizará brevemente los efectos de modificar los operadores, probabilidades de cruce y mutación. Las conclusiones dan un conjunto de pautas aplicables a aquellos investigadores que hagan una primera aproximación al mundo de la evolución gramatical.

09:20
A Fractal-based Mutation Operator from Iterated Function Systems
SPEAKER: unknown

ABSTRACT. In this paper we present a new mutation procedure for Evolutionary Programming (EP) approaches, based on Iterated Function Systems (IFSs). The new mutation procedure proposed consists of considering a set of IFSs, able to generate fractal structures in a two-dimensional phase space, and use them to modify a current individual of the EP algorithm. In other words, the fractal structures generated with the IFSs are randomly selected and sampled from an initial point, obtaining two-dimensional points which will be used to modify the current solution of the EP, instead of using random numbers from different probability density functions. We will describe in detail how to implement this new IFS-M exploration of the search space, and its main characteristics, advantages and disadvantages in terms of the optimization capabilities and implementation properties. We test this new proposal in a real application of Tuned Mass Dumper (TMD) location and optimization for vibration cancellation in buildings. We show how the proposed EP with the IFS-based mutation obtained extremely competitive results compared to alternative classical mutation operators.

09:40
Redes de sensores seguras y eficientes con beamforming

ABSTRACT. Mejorar la eficiencia energética de las redes de sensores (WSN, por sus siglas en inglés) es uno de sus principales objetivos de diseño. De entre las distintas líneas de trabajo existentes en esta línea, el uso del conformado del haz, o beamforming, está en auge en la actualidad, ya que proporciona una forma de transmitir señales de radio de forma muy eficiente hacia un conjunto dado de direcciones destino. En este trabajo, el beamforming se ha utilizado para incrementar, por una parte, el tiempo de vida de las WSNs y, por otro, el nivel de seguridad de la red, evitando establecer comunicaciones en direcciones donde se conoce la existencia de nodos enemigos. El problema se ha formulado agregando dichos objetivos en una única función de fitness, y cuyas soluciones tentativas están compuestas por las amplitudes y fases de las antenas instaladas en los nodos de la WSN. Sobre distintos escenarios sintéticos, los resultados han mostrado que es posible reducir el consumo energético de una WSN y, a la vez, proporcionar comunicaciones seguras ante la presencia de posibles atacantes.

10:00-11:30 Session Plenary: Panel Discussion in Industrial Applications

Mauricio Resende, Principal Research Scientist in the transportation area at Amazon, USA.

Nuria Oliver, Data Science Director of Vodafone and the Chief Data Scientist at Data-Pop Alliance.

Ricardo Baeza-Yates, CTO of NTENT, a semantic search technology company based in California, USA.

Henning Wagner, Head of IT Service Management at Rhenus Freight Logistics.

Kristof Roomp, Principal Software Architect working at Microsoft.

Enrique Alfonseca, Staff Research Scientist at Google.

Chairs:
Location: 30.S01 Auditorium
11:30-12:00 Session : Coffee Break
Location: Patio Jaume I building
12:00-13:30 Session F2.1: Scheduling
Location: 20.017
12:00
A Comparison of Genetic Algorithm and Local Search Based Algorithms for Scheduling Recurring Radiotherapy Treatment Activities with Time Window Constraints and Optional Activities
SPEAKER: Petra Vogl

ABSTRACT. The Radiotherapy Patient Scheduling Problem (RPSP) deals with the assignment of treatment appointments to patients suffering from various types of cancer. The treatment phase consists of recurring irradiation appointments which are followed by optional imaging and examination appointments, all of which require multiple resources simultaneously. The irradiation appointments need to be scheduled at least four times within five days starting from the first appointment, while the optional activities need to take place at least once in every five days. Additionally, we strive for the activity starting times to remain stable within one week. \\We designed two different constructive heuristics: One sequential heuristic and one heuristic which allows filling gaps on idle resources. Further we developed a genetic algorithm approach which works on a multi-encoded chromosome. This encoding scheme allows to assure feasibility during the evolutionary processes of crossover and mutation. Additionally, the multi-encoding scheme is further used within a neighborhood search algorithm. Finally, a hybrid version of the two proposed metaheuristic algorithms is presented and the results of the different approaches are compared on real-life inspired instances.

12:20
A GRASP ALGORITHM FOR SCHEDULING EMERGENCY ROOM PHYSICIANS
SPEAKER: Pedro Mateo

ABSTRACT. This presentation deals with the problem of scheduling emergency room (ER) physicians. The ER never close, it is open 24 h per day and 7 days a week. Physicians are required to work differ-ent shifts, with different assignments in the ER and to work as much on nights and on weekend as during the day. In addition, the consideration of several “ergonomic” type constraints makes the problem of finding good and equitable schedules for the physicians a very difficult problem. It can be modelled as an integer mathematical linear problem, but its dimension prevents from reaching optimal solutions. Then a heuristic approach seems to be the best way to solve it. In this presentation we propose a GRASP type algorithm which is able to get to very good solutions. It is illustrated with a real case in which the schedule of the 43 physicians working in the ER of the Hospital of Navarre for the whole year is obtained. Results are in the process of being tested and validated by the physician in charge of designing the ER-physicians schedules

12:40
An Enhanced Iterated Greedy Metaheuristic for the Particle Therapy Patient Scheduling Problem

ABSTRACT. The Particle Therapy Patient Scheduling Problem (PTPSP) arises in modern cancer treatment facilities that provide particle therapy and consists of scheduling a set of therapies within a planning horizon of several months. A particularity of PTPSP compared to classical radiotherapy scheduling is that therapies need not only be assigned to days but also scheduled within each day to account for the more complicated operational scenario. In an earlier work we introduced this novel problem setting and provided first algorithms including an Iterated Greedy (IG) metaheuristic. In this work we build upon this IG and exchange two main components: the construction phase and the local search algorithm. The resulting metaheuristic enhances the existing approach and yields in most of the considered benchmark instances substantially better results.

13:00
Tabu search and matheuristic algorithms for solving an integrated flow shop and vehicle routing problem

ABSTRACT. We consider in this paper an integrated scheduling and vehicle routing problem: jobs have to be produced and then transported to a customer place for a given due date. The manufacturing workshop is a flow shop and there is only one vehicle available. We are interested in the minimization of the total tardiness, related to the delivery completion times. Therefore, we are faced to three interdependent problems: scheduling the jobs in the flow shop environment, batching the completed jobs, and routing the batches. We propose two tabu search algorithms and a matheuristic algorithm. Computational experiments are performed on random data sets and show the efficiency of the methods.

12:00-13:30 Session F2.2: Air Transportation and Airport Logistics / Smart Cities
Location: 20.019
12:00
Enabling Low Cost Smart Road Traffic Sensing
SPEAKER: Jamal Toutouh

ABSTRACT. Accurate traffic monitoring is a key aspect to develop Smart Mobility services. The bases for monitoring a road traffic system are introduced in this study. The monitoring is done via a network of sensors capturing together information from wireless interfaces of the devices inside the vehicles and also data about noise level. This type of sensors obtain accurate road traffic flows and allow noise maps, that can be further analyzed to provide advanced services for mobility and pollution in the roads. Our main goal here is to develop such sensors at low cost with commodity devices, so as to boost their utilization.

12:20
Internet of Things in urban waste collection
SPEAKER: unknown

ABSTRACT. Nowadays, the waste collection management has an important role in urban areas. This paper faces this issue and proposes the application of a metaheuristic for the optimization of a weekly schedule and routing of the waste collection activities in an urban area. Differently to several contributions in literature, fixed periodic routes are not imposed. The results significantly improve the performance of the company involved, both in terms of resources used and costs saving.

12:40
A simheuristic approach for solving the Aircraft Recovery Problem with stochastic delays
SPEAKER: Wenjing Zhao

ABSTRACT. Air transport operational disruptions arise when operations deviate from the original plan. Due to airlines network configuration, delays are rapidly propagated to connecting flights, substantially increasing unexpected costs for the airlines. The goal in these situations is therefore to minimise the impact of the disruption, reducing delays and the number of affected flights, crews and passengers. However, the reach of a specific disruption is not normally known or it is difficult to assess, increasing the complexity of the problem. In this work, we introduce a methodology based on a Large Neighbourhood Search metaheuristic and a Constraint Programming formulation to tackle the Aircraft Recovery Problem with stochastic delays. We use simulation to help guiding the search, account for system’s variability, and evaluate solutions’ behaviour. We present some preliminary results on a set of instances with different sizes and characteristics, including some instances originating from real data.

13:00
A Simple and Efficient Metaheuristic for the Dynamic Flight Scheduling Problem

ABSTRACT. One of the biggest problems in the airline industry is facing the uncertainty created by passengers demand. In this work, we introduce a metaheuristic that uses updated demand forecasts to optimise a given flight schedule. The algorithm uses two main techniques to solve the problem: (i) retiming, aimed at changing the scheduled time of a flight by delaying or advancing it to allow new connections; and, (ii) refleeting is focused on swapping aircraft between flights to increase aicraft's occupancy and better adjust it to demand. The algorithm's objective is to maximise airlines revenue by increasing the occupancy of each flight. This metaheuristic has been tested in different scenarios of up to 600 flights with promising results in low CPU times.

12:00-13:30 Session F2.3: Metaheuristics - General
Location: 20.021
12:00
Heuristics for the Dynamic Bipartite Drawing Problem

ABSTRACT. Drawings of graphs have many applications and they are nowadays a well established tool in computer science in general and optimization in particular. Project scheduling is one of the many areas in which representation of graphs constitute an important instrument. The experience shows that the main quality desired for drawings of graphs is readability, and crossing reduction is a fundamental aesthetic criterion to achieve it. Incremental graph drawing is an emerging topic in this context, where we seek to preserve the layout of a graph over successive drawings. In this paper we target the edge crossing reduction in the context of incremental graph drawing. Specifically, we apply a mathematical programming formulation and several heuristic methods, mainly based on Path Relinking implementations. In line with a previous paper, we consider bipartite graphs in our experimentation, which shows the superiority of our proposals.

12:20
Computing Short Edge-Flipping Sequences Between Triangulations: a Heuristic Approach

ABSTRACT. In this paper, we propose practical heuristics for the problem of computing the Edge-Flipping Distance Between Triangulations (EFDTP). In [3], Hanke et al. established an upper bound for the problem, along with a high-level algorithm that computes a solution with that bound. We extended that algorithm with a randomized greedy approach, proposed a new greedy strategy and implemented them in order to compare their performance. Extensive computational tests carried out on 240 instances of various sizes and classes revealed that the original algorithm was effective at obtaining good solutions for smaller instances. However, the randomized heuristic in conjunction with the novel greedy strategy lead to improved solution quality for larger cases. Our contribution includes the first experimental analysis of six increasingly efficacious methods for solving the EFDTP.

12:40
On the minimum quartet tree cost problem

ABSTRACT. Given a set of n data objects and their pairwise dissimilarities, the goal of the minimum quartet tree cost (MQTC) problem is to construct an optimal tree from the total number of possible combinations of quartet topologies on n, where optimality means that the sum of the dissimilarities of the embedded (or consistent) quartet topologies is minimal. We provide details and formulation of this novel challenging problem, and the preliminaries of an exact algorithm under current development which may be useful to improve the MQTC heuristics to date into more efficient hybrid approaches.

13:00
Equitable Coloring of Graphe: New genetic method

ABSTRACT. The Graph Coloring Problem has many variations and it is one of the most studied problems in optimization. Various methods have been proposed to solve the classical coloring problem, but few methods in the literature relate to the resolution of the equitable graph coloring problem. An equitable coloring of a graph is defined as a partition of its vertices in k subsets, where the difference on the cardinality of any two subsets is at most one. The Equitable Coloring Problem (ECP) tries to minimize the value of k then a graph can be colored with k colors. We propose in this direction a heuristic balancing that can turn any feasible solution from classical coloration to an equitable coloration solution. We are also inspired from a genetic algorithm for the classic coloring to suggest a genetic approach for the equitable coloring. We have obtained through a broad benchmark of literature interesting results and widely competitive with existing methods.

12:00-13:30 Session F2.4: Metaheuristics - General
Location: 20.023
12:00
HyperSpark: A Software Engineering Approach to Parallel Metaheuristics

ABSTRACT. Metaheuristics are used to solve complex, untractable problems for which other approaches are unsuitable or unable to provide solutions in reasonable times. Although computing power has grown exponentially with the onset of Cloud Computing and Big Data, the domain of metaheuristics has not yet taken full advantage of this new potential. In this paper we address this gap by proposing HyperSpark, a metaheuristic optimization framework for the scalable execution of user-defined, computationally-intensive metaheuristics. HyperSpark provides a way to harness the benefits (e.g., scalability by design) and features (e.g., a simple programming model or ad-hoc infrastructure tuning) of state-of-the-art big data technology for the benefit of metaheuristic computation. We elaborate on HyperSpark and evaluate its efficiency and generality on several different metaheuristics for the Permutation Flow-Shop Problem (PFSP). We observe that HyperSpark results are comparable with the best solutions found and this shows clearly the great potential behind the propsed approach.

12:20
A Multi-Start Algorithm with Intelligent Neighborhood Selection
SPEAKER: Julian Molina

ABSTRACT. In any combinatorial optimization problem, problem-specific solution encodings, such as ordered lists, are usually necessary in order to reduce the complexity of the problem and to provide dynamism whilst obtaining solutions more quickly. A well-known combinatorial optimization problem is the Vehicle Routing Problem, and even more if we need to perform multiple trips because of the capacity of the vehicles is limited and the fleet is reduced. To accomplish this task, an intuitive and easy solution representation is proposed. The solution representation allows encoding a route solution providing to the algorithm dynamism whilst obtaining solutions more quickly at once. This representation is used to solve the Multi-Objective Capacitated Vehicle Routing Problems with Multiple Trips (MO-MTCVRP) where the objectives under consideration are: minimization of the number of vehicles, of the total travel cost, and of the maximum latency (accumulate visiting time for the last node). We have focused in the MO-MTCVRP since many real-world situations fit into this model. For instance, in medical transport operations or in humanitarian logistics activities the objective is to provide a quick service within the minimum possible travel cost but with a reduce fleet of vehicles. For the purpose of producing high-quality solutions, a Multi-Start Algorithm with Intelligent Neighborhood Selection (MSINS) is implemented. This algorithm constructs solutions by using the new representation and is based on the idea of use many different neighborhoods to improve the constructed solution but including some self-adaptive parameters to select the neighborhoods. The MSINS is compared with one of the reference in the literature, the NSGA-II, in order to prove its superiority.

12:40
On the comparison of CMSA versus LNS for solving Combinatorial Optimization problems with different solution sizes

ABSTRACT. Both, Construct, Merge Solve and Adapt (CMSA) and Large Neighborhood Search (LNS), are hybrid algorithms that are based on iteratively solving sub-instances of the original problem instances, if possible, to optimality. This is done by reducing the search space of the tackled problem instance in algorithm-specific ways which differ from one technique to the other. In this paper we provide first experimental evidence for the intuition that, conditioned by the way in which the search space is reduced, LNS should generally work better than CMSA in the context of problems in which solutions are rather large, and the opposite is the case for problems in which solutions are rather small. The size of a solution is hereby measured by the number of components of which the solution is composed, in comparison to the total number of solution components. In this ongoing work we are conducting experiments in the context of the multi-dimensional knapsack problem, the minimum-weight dominating set, and the single-source capacitated facility location problem.

13:00
Extending time-to-target plots to multiple instances and targets: mttt-plots
SPEAKER: Alberto Reyes

ABSTRACT. Time-to-target plots (tttplots) are a useful tool to characterize, evaluate, and compare the behavior of randomized heuristics for a given problem instance of some combinatorial optimization problem. Multiple time-to-target plots (m-tttplots) are their natural extension to sets of multiple instances. We show how to build an m-tttplots from the individual tttplots for each instance in the set. Finally, we give an illustrative example of the construction of the mttt-plots from the individual ttt-plots for a case study involving GRASP algorithms for the 2-path network design problem.

12:00-13:30 Session F2.5: Metaheuristics General (MAEB)
Location: 20.025
12:00
Un GRASP/VNS para un problema de recogida de leche modelado como PVRPTW

ABSTRACT. Los modelos asociados a los problemas de recolección y distribución de productos perecederos deben adoptar las especificidades operacionales de los sistemas de logística y transporte que trabajan con este tipo de productos. Estos modelos de optimización deben tender a asegurar la calidad del servicio y de los productos perecederos. En este trabajo se aborda un problema de planificación de recolección de leche fresca en granjas con instalaciones isotérmicas limitadas para el mantenimiento del producto. Esta actividad demanda una organización temporal para preservar su calidad por un periodo al menos semanal. Específicamente, el problema de determinar el conjunto de rutas más apropiadas para recolectar leche fresca de un conjunto de granjas se formula como un PVRPTW (Periodic Vehicle Routing Problem with Time Windows). Para su solución se diseña un algoritmo híbrido que integra una VND como fase de optimización de un GRASP.

12:20
Nuevas heurísticas de construcción para el problema del ordenamiento óptimo de buckets
SPEAKER: Jose Gamez

ABSTRACT. El método de Borda es una forma muy usual de resolver el problema de Kemeny, que pretende encontrar la permutación que tiene menos desacuerdos con conjunto de permutaciones de entrada. El problema del ordenamiento óptimo de buckets (OBOP), es una variante del problema de Kemeny, donde las soluciones son rankings completos en los que se admiten empates entre ítems. En este trabajo se presentan y estudian varias heurísticas para resolver el OBOP que se basan en estrategias de ordenar primero y agrupar después, y que emplean el método de Borda en la etapa inicial (ordenar). Los resultados preliminares muestran una ganancia del 20% sobre el método voraz habitualmente utilizado, BPA (Bucket Pivot Algorithm), lo que indica que estas heurísticas son excelentes candidatas para ser usadas en las etapas de inicialización para la solución mediante metaheurísticas del OBOP.

12:40
Un algoritmo genético para resolver el problema de se-cuenciación en carruseles con huecos
SPEAKER: Carlos Andres

ABSTRACT. En este trabajo se presenta un problema de secuenciación en carruseles en los que pueden aparecer pérdidas de capacidad debido a la necesidad de varios tipos de preparación. Este problema ha sido estudiado previamente considerando que estas pérdidas suponen un coste. Sin embargo, la consideración de huecos en la secuencia permiten modelar el sistema productivo de forma más realista. Junto con el modelo matemático se plantea un algoritmo genético destinado a resolver el problema para instancias del problema realistas. En el trabajo no se detallan los resultados del algoritmo que sí que se presentarán en el congreso MIC/MAEB 2017.