The benchmark functions and some of the algorithms proposed for the special session on real parameter optimization of the 2005 IEEE Congress on Evolutionary Computation (CEC'05) have played and still play an important role in the assessment of the state of the art in continuous optimization. In this note, we first show that, if bound constraints are not enforced for the final reported solutions, state-of-the-art algorithms produce on a majority of the CEC'05 benchmark functions infeasible best candidate solutions. This is the case even though the optima of the CEC'05 functions are within the specified bounds. This observation has important implications on algorithm comparisons and this note also draws the attention to the fact that authors may have drawn wrong conclusions from experiments using the CEC'05 problems.

@article{LiaMonStu12,
author = {Tianjun Liao and Daniel Molina and Marco A. {Montes de Oca} and Thomas St{\"u}tzle},
title = {A Note on Bound Constraints Handling for the IEEE CEC'05 Benchmark Function Suite. Evolutionary Computation},
journal = {Evolutionary Computation},
year = {2014},
volume = {22},
number = {2},
pages = {351--359}
}

In this article, we propose UACOR, a unified ant colony optimization (ACO) algorithm for continuous optimization. UACOR includes algorithmic components from ACOR, DACOR and IACOR-LS, three ACO algorithms for continuous optimization that have been proposed previously. Thus, it can be used to instantiate each of these three earlier algorithms; in addition, from UACOR we can also generate new continuous ACO algorithms that have not been considered before in the literature. In fact, UACOR allows the usage of automatic algorithm configuration techniques to automatically derive new ACO algorithms. To show the benefits of UACOR's flexibility, we automatically configure two new ACO algorithms, UACOR-s and UACOR-c, and evaluate them on two sets of benchmark functions from a recent special issue of the Soft Computing (SOCO) journal and the IEEE 2005 Congress on Evolutionary Computation (CEC'05), respectively. We show that UACOR-s is competitive with the best of the 19 algorithms benchmarked on the SOCO benchmark set and that UACOR-c performs superior to IPOP-CMA-ES and statistically significantly better than five other algorithms benchmarked on the CEC'05 set. These results show the high potential ACO algorithms have for continuous optimization and suggest that automatic algorithm configuration is a viable approach for designing state-of-the-art continuous optimizers.

@article{LiaMonStu12,
author = {Tianjun Liao and Thomas St{\"u}tzle and Marco A. {Montes de Oca} and Marco Dorigo},
title = {A Unified Ant Colony Optimization Algorithm for Continuous Optimization},
journal = {European Journal of Operational Research},
year = {2014},
volume = {243},
number = {3},
pages={597--609}
}

In this paper, we introduce ACOMV, an ant colony optimization (ACO) algorithm that extends the ACOR algorithm for continuous optimization to tackle mixed-variable optimization problems. In ACOMV, the decision variables of an optimization problem can be explicitly declared as continuous, ordinal, or categorical, which allows the algorithm to treat them adequately. ACOMV includes three solution generation mechanisms: a continuous optimization mechanism (ACOR), a continuous relaxation mechanism (ACOMV-o) for ordinal variables, and a categorical optimization mechanism (ACOMV-c) for categorical variables. Together, these mechanisms allow ACOMV to tackle mixedvariable optimization problems. We also define a novel procedure to generate artificial, mixed-variable benchmark functions and we use it to automatically tune ACOMV's parameters. The tuned ACOMV is tested on various real-world continuous and mixed-variable engineering optimization problems. Comparisons with results from the literature demonstrate the effectiveness and robustness of ACOMV on mixed-variable optimization problems.

@article{LiaMonStu12,
author = {Tianjun Liao and Krzysztof Socha and Marco A. {Montes de Oca} and Thomas St{\"u}tzle and Marco Dorigo},
title = {Ant Colony Optimization for Mixed-Variable Optimization Problems},
journal = {IEEE Transactions on Evolutionary Computation},
year = {2014},
volume = {18},
number = {4},
pages = {503--518}
}

In this article, we apply an automatic algorithm configuration tool to improve the performance of the CMA-ES algorithm with increasing population size (iCMA-ES), the best performing algorithm on the CEC'05 benchmark set for continuous function optimization. In particular, we consider a separation between tuning and test sets and, thus, tune iCMA-ES on a different set of functions than the ones of the CEC'05 benchmark set. Our experimental results show that the tuned iCMA-Es improves significantly over the default version of iCMA-ES. Furthermore, we provide some further analysis on the impact of the modified parameter settings on iCMA-ES performance and a comparison to recent results of algorithms that use CMA-ES as a subordinate local search.

@article{LiaMonStu12,
author = {Tianjun Liao and Marco A. {Montes de Oca} and Thomas St{\"u}tzle},
title = {Computational Results for an Automatically Tuned {CMA}-{ES} with Increasing Population Size on the {CEC}'05 Benchmark Set},
journal = {Soft Computing},
year = {2013},
volume = {17},
number = {6},
pages={1031--1046}
}

The problem of inferring the structure of a network of interactions between biological variables from a set of observations is called reverse engineering. In this paper, we propose an optimization algorithm, called MORE, that tackles the two components of this problem: a discrete component, which represents the topology of the underlying molecular interaction network, and a continuous component, which represents the strength of the interactions. The model used to generate candidate time series patterns is a sparse system of nonlinear differential equations, complex enough to realistically describe the dynamics of a biological system. The goal of the optimization process is to minimize the error between the model’s dynamics and the observed data. Decomposing the problem into the two aforementioned components allows us both to enforce system sparsity, by globally constraining the number of edges in the network, and to easily integrate a priori information about the structure of the underlying interaction network. The design of MORE is based on thorough experimentation with simulated and real-world networks of up to 10 nodes, which is larger than the size of the networks used in the majority of the current proposals of reverse engineering algorithms based on sufficiently complex interaction models.

@article{SamMonDiCTofStu12,
author = {Francesco Sambo and Marco A. {Montes de Oca} and Barbara {Di Camillo} and Giovanna Toffolo and Thomas St{\"u}tzle},
title = {{MORE}: {M}ixed {O}ptimization for {R}everse {E}ngineering. An application to modeling biological networks response via sparse systems of nonlinear differential equations},
journal = {IEEE/ACM Transactions on Computational Biology and Bioinformatics},
year = {2012},
volume = {9},
number = {5},
pages={1459--1471}
}

The performance of optimization algorithms, including those based on swarm intelligence, depends on the values assigned to their parameters. To obtain high performance, these parameters must be fine-tuned. Since many parameters can take real values or integer values from a large domain, it is often possible to treat the tuning problem as a continuous optimization problem. In this article, we study the performance of a number of prominent continuous optimization algorithms for parameter tuning using various case studies from the swarm intelligence literature. The continuous optimization algorithms that we study are enhanced to handle the stochastic nature of the tuning problem. In particular, we introduce a new post-selection mechanism that uses F-Race in the final phase of the tuning process to select the best among elite parameter configurations. We also examine the parameter space of the swarm intelligence algorithms that we consider in our study and we show that by fine-tuning their parameters one can obtain substantial improvements over default configurations.

@article{ZhiMonBirStu12,
author = {Yuan Zhi and Marco A. {Montes de Oca} and Mauro Birattari and Thomas St{\"u}tzle},
title = {Continuous Optimization Algorithms for Tuning Real and Integer Parameters of Swarm Intelligence Algorithms},
journal = {Swarm Intelligence},
year = {2012},
volume = {6},
number = {1},
pages={49--75}
}

Swarm robotics systems are characterised by decentralised control, limited communication between robots, use of local information and emergence of global behaviour. Such systems have shown their potential for flexibility and robustness. However, existing swarm robotics systems are by in large still limited to displaying simple proof-of-concept behaviours under laboratory conditions. It is our contention that one of the factors holding back swarm robotics research is the almost universal insistence on homogeneous system components. We believe that swarm robotics designers must embrace heterogeneity if they ever want swarm robotics systems to approach the complexity required of real world systems.

@article{Swarmanoid,
author = {Marco Dorigo and Dario Floreano and Luca Maria Gambardella and Francesco Mondada and Stefano Nolfi and Tarek Baaboura and Mauro Birattari and Michael Bonani and Manuele Brambilla and Arne Brutschy and Daniel Burnier and Alexandre Campo and Anders Lyhne Christensen and Antal Decugni{\`e}re and Gianni Di Caro and Frederick Ducatelle and Eliseo Ferrante and Alexander F{\"o}rster and Javier Martinez Gonzales and Jerome Guzzi and Valentin Longchamp and St{\'e}phane Magnenat and Nithin Mathews and Marco A. {Montes de Oca} and Rehan O’Grady and Carlo Pinciroli and Giovanni Pini and Philippe R{\'e}tornaz and James Roberts and Valerio Sperati and Timothy Stirling and Alessandro Stranieri and Thomas St{\"u}tzle and Vito Trianni and Elio Tuci and Ali Emre Turgut and and Florian Vaussard},
title = {Swarmanoid: A novel concept for the study of heterogeneous robotic swarms},
journal = {IEEE Robotics \& Automation Magazine},
year = {2013},
volume = {20},
number = {4},
pages={60--71}
}

Collective decision-making is a process whereby the members of a group decide on a course of action by consensus. In this paper, we propose a collective decision-making mechanism for robot swarms deployed in scenarios in which robots can choose between two actions that have the same effects but that have different execution times. The proposed mechanism allows a swarm composed of robots with no explicit knowledge about the difference in execution times between the two actions to choose the one with the shorter execution time. We use an opinion formation model that captures important elements of the scenarios in which the proposed mechanism can be used in order to predict the system's behavior. The model predicts that when the two actions have different average execution times, the swarm chooses with high probability the action with the shorter average execution time. We validate the model's predictions through a swarm robotics experiment in which robots must choose one of two paths of different length that connect two locations. Thanks to the proposed mechanism, a swarm made of robots that do not measure time or distance is able to choose the shorter path.

@article{MonFerSchPinBirDor11,
author = {Marco A. {Montes de Oca} and Eliseo Ferrante and Alexander Scheidler and Carlo Pinciroli and Mauro Birattari and Marco Dorigo},
title = {Majority-Rule Opinion Dynamics with Differential Latency: A Mechanism for Self-Organized Collective Decision-Making},
journal = {Swarm Intelligence},
year = {2011},
volume = {5},
number = {3--4},
pages={305--327}
}

The design cycle of high-performance optimization algorithms requires the designer to make several decisions. These decisions range from implementation details to the setting of parameter values for testing intermediate designs. Proper parameter setting can be crucial for the effective assessment of algorithmic components because a bad parameter setting can make a good algorithmic component perform poorly. This situation may lead the designer to discard promising components that just happened to be tested with bad parameter settings. Automatic parameter tuning techniques are being used by practitioners to obtain peak performance from already designed algorithms. However, automatic parameter tuning also plays a crucial role during the design cycle of optimization algorithms. In this paper, we present a case study of a tuning-in-the-loop approach for redesigning a particle swarm-based optimization algorithm for tackling large-scale continuous optimization problems. Rather than just presenting the final algorithm, we describe the whole redesign process. Lastly, we study the scalability behavior of the final algorithm in the context of this special issue.

@article{MonAydStu11,
author = {Marco A. {Montes~de~Oca} and Dogan Aydin and Thomas St{\"u}tzle},
title = {An Incremental Particle Swarm for Large-Scale Continuous Optimization Problems: An Example of Tuning-in-the-loop (Re)Design of Optimization Algorithms},
journal = {Soft Computing},
volume = {15},
number = {11},
pages={2233--2255},
year = {2011} }

Version: 1.0
Author: Marco A. Montes de Oca, Dogan Aydin, and Thomas Stützle
Copyright (c) Marco A. Montes de Oca, Dogan Aydin, and Thomas Stützle, 2010
Download

Incremental social learning (ISL) was proposed as a way to improve the scalability of systems composed of multiple learning agents. In this article, we show that ISL can be very useful to improve the performance of population-based optimization algorithms. Our study focuses on two particle swarm optimization (PSO) algorithms: a) the incremental particle swarm optimizer (IPSO), which is a PSO algorithm with a growing population size in which the initial position of new particles is biased toward the best-so-far solution, and b) the incremental particle swarm optimizer with local search (IPSOLS), in which solutions are further improved through a local search procedure.

We first derive analytically the probability density function induced by the proposed initialization rule applied to new particles. We then compare the performance of IPSO and IPSOLS on a set of benchmark functions with that of other PSO algorithms (with and without local search) and a random restart local search algorithm. Finally, we measure the benefits of using incremental social learning on PSO algorithms by running IPSO and IPSOLS on problems with different fitness distance correlations.

@article{ISLPSMdeO,
author = {Marco A. {Montes~de~Oca} and Thomas St{\"u}tzle and Ken {Van den Enden} and Marco Dorigo},
title = {Incremental Social Learning in Particle Swarms},
journal = {IEEE Transactions on Systems, Man, and Cybernetics - Part B: Cybernetics},
volume={41},
number={2},
pages={368--384},
year={2011}
}

During the last decade, many variants of the original particle swarm optimization (PSO) algorithm have been proposed. In many cases, the difference between two variants can be seen as an algorithmic component being present in one variant but not in the other. In the first part of the paper, we present the results and insights obtained from a detailed empirical study of several PSO variants from a component difference point of view. In the second part of the paper, we propose a new PSO algorithm that combines a number of algorithmic components that showed distinct advantages in the experimental study concerning optimization speed and reliability. We call this composite algorithm Frankenstein's PSO in an analogy to the popular character of Mary Shelley's novel. Frankenstein's PSO performance evaluation shows that by integrating components in novel ways effective optimizers can be designed.

@article{FPSOMdeO09,
author = {Marco A. {Montes~de~Oca} and Thomas St{\"u}tzle and Mauro Birattari and Marco Dorigo},
title = {Frankenstein's {PSO}: {A} Composite Particle Swarm Optimization Algorithm},
journal = {IEEE Transations on Evolutionary Computation},
volume = {13},
number = {5},
pages={1120--1132},
year = {2009}
}

Version: 1.0
Author: Marco A. Montes de Oca, Thomas Stützle, Mauro Birattari, and Marco Dorigo
Copyright (c) Marco A. Montes de Oca,Thomas Stützle, Mauro Birattari, and Marco Dorigo, 2010
Download

Particle swarm optimization is a population-based stochastic approach for solving continuous and discrete optimization problems.

In particle swarm optimization, simple software agents, called particles, move in the solution space of an optimization problem. The position of a particle represents a candidate solution to the optimization problem at hand. Particles search for better positions in the solution space by changing their velocity according to rules originally inspired by behavioral models of bird flocking.

Particle swarm optimization belongs to the class of swarm intelligence techniques that are used to solve optimization problems.

@article{PSODorigo08,
author = "Marco Dorigo and Marco A. {Montes~de~Oca} and Andries P. Engelbrecht",
title = "Particle Swarm Optimization",
journal = "Scholarpedia",
volume = "3",
number = "11",
pages = "1486",
year = "2008"
}

Los algoritmos de agrupación de clases basados en el comportamiento colectivo de algunas especies de insectos sociales son herramientas para el descubrimiento de conocimiento que se basan en la comunicación estigmérgica, esto es, en la comunicación indirecta entre agentes a través de modificaciones locales al ambiente. Inspirados en algunos aspectos del fenómeno de la trofalaxis, o el intercambio de alimento líquido entre insectos, estudiamos dos estrategias de comunicación directa entre agentes en este tipo de algoritmos: (i) compartición de memoria y (ii) compartición de mapas ambientales. El objetivo del intercambio de información entre agentes es mejorar la calidad de la agrupación final. El efecto de la inclusión de estos mecanismos de comunicación se evalúa mediante la comparación de la calidad de la agrupación final obtenida con comunicación directa y sin comunicación (aparte de la estigmérgica). Se muestra además, que los resultados dependen de la densidad de agentes en el ambiente, así como de la utilidad de la información intercambiada.

@article{ComDirectMdeO05,
author = "Marco A. {Montes~de~Oca} and Leonardo Garrido and Jos\'e L. Aguirre",
title = "Efectos de la comunicaci\'on directa entre agentes en los algoritmos de agrupaci\'on de clases basados en el comportamiento de insectos sociales",
journal = "Inteligencia Artificial",
volume = "9",
number = "25",
pages = "59--69",
year = "2005"
}

Swarm intelligence is the collective problem-solving behavior of groups of animals and artificial agents. Often, swarm intelligence is the result of self-organization, which emerges from the agents' local interactions with one another and with their environment. Such local interactions can be positive, negative, or neutral. Positive interactions help a swarm of agents solve a problem. Negative interactions are those that block or hinder the agents' task-performing behavior. Neutral interactions do not affect the swarm's performance. Reducing the effects of negative interactions is one of the main tasks of a designer of effective swarm intelligence systems. Traditionally, this has been done through the complexification of the behavior and/or the characteristics of the agents that comprise the system, which limits scalability and increases the difficulty of the design task. In collaboration with colleagues, I have proposed a framework, called incremental social learning (ISL), as a means to reduce the effects of negative interactions without complexifying the agents' behavior or characteristics. In this paper, I describe the ISL framework and three instantiations of it, which demonstrate the framework's effectiveness. The swarm intelligence systems used as case studies are the particle swarm optimization algorithm, ant colony optimization algorithm for continuous domains, and the artificial bee colony optimization algorithm.

@InCollection{Mon12,
author = {Marco A. {Montes~de~Oca}},
title = {Incremental Social Learning in Swarm Intelligence Algorithms for Continuous Optimization},
booktitle = {Computational Intelligence. Revised and Selected Papers of the International Joint Conference, IJCCI 2011.},
publisher = {Springer},
pages = {31--45},
address = {Berlin, Germany},
year = 2013,
editor = {Kurosh Madani and others}
}

In memetic algorithms, the improvement of individual solutions through local search is of crucial importance for attaining high performance. In this chapter, we give a concise overview of the main issues that arise in the definition of local search algorithms, their implementation and their integration into a memetic algorithm. Our overview covers local search algorithms for both combinatorial and continuous optimization problems.

@InCollection{MonCotNer11,
author = {Marco A. {Montes~de~Oca} and Carlos Cotta and Ferrante Neri},
title = {Local Search},
booktitle = {Handbook of Memetic Algorithms},
publisher = {Springer},
pages = {29--41},
address = {Berlin, Germany},
year = 2012,
editor = {F. Neri and C. Cotta and P. Moscato}
}

This chapter reviews the approaches that have been studied for the online adaptation of the parameters of ant colony optimization (ACO) algorithms, that is, the variation of parameter settings while solving an instance of a problem. We classify these approaches according to the main classes of online parameter-adaptation techniques. One conclusion of this review is that the available approaches do not exploit an in-depth understanding of the effect of individual parameters on the behavior of ACO algorithms. Therefore, this chapter also presents results of an empirical study of the solution quality over computation time for Ant Colony System and MAX-MIN Ant System, two well-known ACO algorithms. The first part of this study provides insights on the behaviour of the algorithms in dependence of fixed parameter settings. One conclusion is that the best fixed parameter settings of MAX-MIN Ant System depend strongly on the available computation time. The second part of the study uses these insights to propose simple, pre-scheduled parameter variations. Our experimental results show that such pre-scheduled parameter variations can dramatically improve the anytime performance of MAX-MIN Ant System.

@InCollection{Stuetzle11PAACO,
author = {Thomas St{\"u}tzle and Manuel L{\'o}pez-Ib{\'a}{\~n}ez and Paola Pellegrini and Michael Maur and Marco A. {Montes~de~Oca} and Mauro Birattari and Marco Dorigo},
title = {Parameter Adaptation in Ant Colony Optimization},
booktitle = {Autonomous Search},
pages = {191--215},
publisher = {Springer},
address = {Berlin, Germany},
year = 2011,
editor = {Y. Hamadia and E. Monfroy and F. Saubion},
}

Ant Colony Optimization (ACO) is a class of algorithms for tackling optimization problems that is inspired by the pheromone trail laying and following behavior of some ant species. While foraging, ants leave on the ground a chemical substance, called pheromone, that attracts other fellow nestmates. The pheromone trail laying and following behavior of the ants induces a positive feedback process whereby trails with high concentration of pheromones become more and more attractive as more ants follow them. As a result, whenever two paths to the same food source are discovered, the colony is more likely to select the shortest one because ants will traverse it faster and thus it will have a higher pheromone concentration than the longer one.

ACO algorithms exploit a mechanism analogous to the one that allows colonies of real ants to find shortest paths. In ACO, (artificial) ants construct candidate solutions to the problem instance under consideration. Their solution construction is stochastically biased by (artificial) pheromone trails, which are represented in the form of numerical information that is associated with appropriately defined solution components, and possibly by heuristic information based on the input data of the instance being solved. A key aspect of ACO algorithms is the use of a positive feedback loop implemented by iterative modifications of the artificial pheromone trails that are a function of the ants' search experience; the goal of this feedback loop is to bias the colony towards the most promising solutions.

The ACO metaheuristic is a high-level algorithmic framework for applying the above ideas to the approximate solution of optimization problems. When applied to a specific optimization problem, this ACO framework needs to be concretized by taking into account the specifics of the problem under consideration and possibly by adding additional techniques such as problem-specific solution improvement procedures. The development of effective ACO algorithm variants has been one of the most active research directions in ACO. This chapter gives an overview of some of the most important developments in this direction.

@incollection{DorMonOliStu2011eorms,
Author = {Marco Dorigo and Marco A. {Montes de Oca} and Sabrina Oliveira and Thomas St{\"u}tzle},
Booktitle = {Wiley Encyclopedia of Operations Research and Management Science},
Editor = {Cochran, James J. and Cox, Louis A. and Keskinocak, Pinar and Kharoufeh, Jeffrey P. and Smith, J. Cole},
Pages = {114--125},
Publisher = {John Wiley \& Sons, Inc.},
Title = {Ant Colony Optimization},
Volume = {1},
Year = {2011}}

In this paper, we present a continuous-time binary consensus protocol whereby entities connected via a directed ring topology solve the one-dimensional density classification problem. In our model, the participating entities behave as non-ideal relays, that is, they have memory of the trajectory of an internal state variable, which gives them hysteretic properties. We show that this feature is necessary for the system to reach consensus on the state shared by the initial majority. The connections between this protocol and collective decision-making mechanisms in swarm intelligence systems are also discussed.

@incollection{MonRos14,
Author = {Marco A. {Montes de Oca} and Louis F.\ Rossi},
Booktitle = {Proceedings of the International Conference on the Synthesis and Simulation of Living Systems (ALIFE 2014)},
Title = {A Continuous-Time Binary Consensus Protocol with Hysteretic Units Applied to the Density Classification Problem},
Editor = {H.\ Sayama and others},
Address={Cambridge, MA},
Publisher={MIT Press},
Notes = {To appear},
Year = {2014}
}

In previous work, we introduced a novel swarming interpolation framework and validated its effectiveness on static fields. In this paper, we show that a slightly revised version of this framework is able to track fields that translate, rotate, or expand over time, enabling interpolation of both static and dynamic fields. Our framework can be used to control autonomous mobile sensors into flexible spatial arrangements in order to interpolate values of a field in an unknown region. The key advantage to this framework is that the stable sensor distribution can be chosen to resemble a Chebyshev distribution, which can be optimal for certain ideal geometries.

@incollection{KirMonSenRosShe13,
Author = {Joshua Kirby and Marco A. {Montes de Oca} and Steven Senger and Louis F.\ Rossi and Chien-Chung Shen},
Booktitle = {Proceedings of the IEEE Conference on Self-Adaptive and Self-Organizing Systems (SASO 2013)},
Title = {Tracking Time-dependent Scalar Fields with Swarms of Mobile Sensors},
Editor = {T.\ Holvoet and others},
Address={Los Alamitos, CA},
Publisher={IEEE Computer Society Press},
Pages = {159--168},
Year = {2013}
}

Automated algorithm configuration methods have proven to be instrumental in deriving high-performing algorithms and such methods are increasingly often used to configure evolutionary algorithms. One major challenge in devising automatic configuration techniques is to handle the inherent stochasticity in the configuration problems. This article analyses a post-selection mechanism that can also be used for this task. The central idea of the post-selection mechanism is to generate ina first phase a set of high-quality candidate algorithm configurations and then to select in a second phase from this candidate set the (statistically) best configuration. Our analysis of this mechanism indicates its high potential and suggests that it may be helpful to improve automatic algorithm configuration methods.

@incollection{YuaStuLauBirMon13,
Author = {Zhi Yuan and Thomas St{\"u}tzle and Hoong Chuin Lau and Mauro Birattari and Marco A. {Montes de Oca}},
Booktitle = {Proceedings of Genetic and Evolutionary Computation Conference (GECCO 2013)},
Title = {An Analysis of Post-Selection in Automatic Configuration},
Editor = {Christian Blum},
Pages = {1557--1564},
Year = {2013},
Address={New York, NY},
Publisher={ACM Press}
}

In this paper, we reinterpret the most basic exponential smoothing equation as a model of social influence. Exponential smoothing has been used as a time-series forecasting and data filtering technique since the 1950s. The most basic exponential smoothing equation, $S^{\,t+1} = (1-\alpha)S^{t} + \alpha X^{t}$, is used to estimate the value of a series at time $t+1$, denoted by $S^{\,t+1}$, as a convex combination of the current estimate $S^{t}$ and the actual observation of the time series $X^{t}$. In our work, we interpret the variable $S^{t}$ as an agent's tendency to imitate the observed behavior of another agent, which is represented by a binary variable $X^{t}$. We study the dynamics of the resulting system when the agents exhibit a behavior for a period of time, called latency, whose duration is stochastic. Latency allows us to model real-life situations such as product adoption, or action execution. When different latencies are associated with the two different behaviors, a bias is produced. This bias makes all the agents in a population adopt one specific behavior. This phenomenon has applications in domains where it is desired that a group of agents adopts a behavior by consensus.

@incollection{MonFerSchRos12,
Author = {Marco A. {Montes de Oca} and Eliseo Ferrante and Alexander Scheidler and Louis F. Rossi},
Booktitle = {LNICST 126. Proceedings of the Second International Conference on Complex Sciences: Theory and Applications (COMPLEX 2012)},
Title = {Binary Consensus via Exponential Smoothing},
Editor = {Kristin Glass and others},
Pages={244--255},
Year = {2013},
Address={Berlin, Germany},
Publisher={Springer}
}

ACO_R is an ant colony optimization algorithm for continuous domains. In this article, we benchmark ACO_R on the BBOB noiseless function testbed, and compare its performance to PSO, ABC and GA algorithms from previous BBOB workshops. Our experiment shows that ACO_R performs better than PSO, ABC and GA on the moderate functions, ill-conditioned functions and multi-modal functions. Among 24 functions, ACO_R solved 19 in dimension 5, 9 in dimension 20, and 7 across dimensions from 2 to 40. Furthermore, in dimension 5, we present the results of the ACO_R when it uses variable correlation handling. The latter version is competitive on the ve dimensional functions to (1+1)-CMA-ES and BIPOP-CMA-ES.

@incollection{LiaMolStuMonDor12,
Author = {Tianjun Liao and Daniel Molina and Thomas St{\"u}tzle and Marco A. {Montes de Oca} and Marco Dorigo},
Booktitle = {Proceedings of the Workshop on Black-Box Optimization Benchmarking of the Genetic and Evolutionary Computation Conference (GECCO 2012)},
Editor = {A. Auger and N. Hansen and V. Heidrich-Meisner and O. Mersmann and P. Posik and M. Preuss},
Title = {An {ACO} Algorithm Benchmarked on the {BBOB} Noiseless Function Testbed},
Year={2012},
Pages = {159--166},
Address={New York, NY},
Publisher={ACM Press}}
}

In this paper, we describe a novel swarming framework that guides autonomous mobile sensors into a flexible arrangement to interpolate values of a field in an unknown region. The algorithm is devised so that the locations of the sensors approximate a Chebyshev distribution, which can be optimal for certain ideal geometries. The framework is designed to dynamically adjust to changes in the region of interest, and operates well with very little a priori knowledge of the given region. For comparison, we interpolate a variety of nontrivial fields using a standard swarming algorithm that produces a uniform distribution and our new algorithm. We find that our new algorithm interpolates fields with greater accuracy.

@incollection{KirMonSenRosShe12,
Author = {Josh Kirby and Marco A. {Montes de Oca} and Steven Senger and Louis F. Rossi and Chien-Chung Shen},
Booktitle = {Proceedings of the Eighth International Conference on Swarm Intelligence (ANTS 2012)},
Title = {Swarm Interpolation Using an Approximate Chebyshev Distribution},
Editor = {M. Dorigo and others},
Pages = {324--331},
Address={Berlin, Germany},
Publisher={Springer},
Year = {2012}
}

We modify an artificial bee colony algorithm to (i) make the population size grow over time, and to (ii) apply local search on strategically selected solutions. The modified algorithm obtains very good results on a set of large-scale continuous optimization problems. This is not the first time we see that these algorithmic components make an initially non-competitive algorithm produce state-of-the-art results. In previous work, we have shown that the same modifications substantially improve the performance of particle swarm optimization (PSO) and ant colony optimization (ACO) algorithms. Altogether, these results suggest that population growth and local search are key components for obtaining high-quality results.

@incollection{AydLiaMonStu11,
Author = {Dogan Aydin and Tianjun Liao and Marco A. {Montes de Oca} and Thomas St{\"u}tzle},
Booktitle = {LNCS 7401. Proceedings of the International Conference on Artificial Evolution (EA 2011)},
Title = {Improving Performance via Population Growth and Local Search: The Case of the Artificial Bee Colony Algorithm},
Pages = {85--96},
Address = {Berlin, Germany},
Publisher = {Springer},
Year={2012}
}

Sep-G-CMA-ES is a variant of G-CMA-ES with lower time complexity. In this paper, we evaluate the impact various ways of tuning have on the performance of Sep-G-CMAES on large scalable continuous benchmark problems. We have extracted seven parameters from Sep-G-CMA-ES and tuned them across training functions with various search landscapes by an automatic algorithm configuration tool, Iterated F-Race. Considering different compositions of training functions, the best performance of tuned Sep-G-CMAES was obtained using mixed dimensional problems. Our comparative study on large scalable benchmark functions also shows that the default Sep-G-CMA-ES outperforms GCMA- ES. Moreover, tuned Sep-G-CMA-ES significantly improves over both G-CMA-ES and default Sep-G-CMA-ES

@incollection{LiaMonStu11,
Author = {Tianjun Liao and Marco A. {Montes de Oca} and Thomas St{\"u}tzle},
Booktitle = {Proceedings of the Workshop on Scaling Behaviours of Landscapes, Parameters and Algorithms of the Genetic and Evolutionary Computation Conference (GECCO 2011)},
Editor = {Ender {\"O}zcan and Andrew J. Parkes and Jonathan Rowe},
Title = {Tuning Parameters across Mixed Dimensional Instances: A Performance Scalability Study of Sep-G-CMA-ES},
Year={2011},
Pages = {703--706},
Address={New York, NY},
Publisher={ACM Press}}
}

ACO_R is one of the most popular ant colony optimization algorithms for tackling continuous optimization problems. In this paper, we propose IACO_R-LS, which is a variant of ACO_R that uses local search and that features a growing solution archive. We experiment with Powell's conjugate directions set, Powell's BOBYQA, and Lin-Yu Tseng's Mtsls1 methods as local search procedures. Automatic parameter tuning results show that IACO_R-LS with Mtsls1 (IACO_R-Mtsls1) is not only a significant improvement over ACO_R, but that it is also competitive with the state-of-the-art algorithms described in a recent special issue of the Soft Computing journal. Further experimentation with IACO_R- Mtsls1 on an extended benchmark functions suite, which includes functions from both the special issue of Soft Computing and the IEEE 2005 Congress on Evolutionary Computation, demonstrates its good performance on continuous optimization problems.

@incollection{LiaMonDogStuDor11,
Author = {Tianjun Liao and Marco A. {Montes de Oca} and Dogan Aydin and Thomas St{\"u}tzle and Marco Dorigo},
Booktitle = {Proceedings of the Genetic and Evolutionary Computation Conference (GECCO 2011)},
Editor = {N. Krasnogor and others},
Title = {An Incremental Ant Colony Algorithm with Local Search for Continuous Optimization},
Year={2011},
Pages = {125--132},
Address={New York, NY},
Publisher={ACM Press}}
}

Positive feedback and a consensus-building procedure are the key elements of a mechanism that allows a population of agents to select the fastest-to-execute action from a set of two alternatives. This self-organized decision-making mechanism can be seen as a collective learning algorithm because even though individual agents do not directly compare the available alternatives, the population selects the action that improves the efficiency of the system. However, when a large population is involved, the time required for achieving consensus on one of the available choices may render impractical such a decision-making mechanism.

In this paper, we tackle this problem by applying the incremental social learning approach, which consists of a growing population size coupled with a social learning mechanism. The experimental results obtained show that by using the incremental social learning approach, the collective learning process can be accelerated substantially. The conditions under which this is true are described.

@incollection{MdeO:ISLSASO10,
Author = {Marco A. {Montes de Oca} and Thomas St{\"u}tzle and Mauro Birattari and Marco Dorigo},
Booktitle = {Proceedings of the Fourth IEEE International Conference on Self-Adaptive and Self-Organizing Systems (SASO 2010)},
Editor = {I. Gupta and others},
Title = {Incremental Social Learning Applied to a Decentralized Decision-Making Mechanism: Collective Learning Made Faster},
Pages={243--252},
Year={2010},
Note = {Preliminary version available at \url{http://iridia.ulb.ac.be/IridiaTrSeries/IridiaTr2010-018r001.pdf}}
}

In this paper, we study how an opinion dynamics model can be the core of a collective decision-making mechanism for swarm robotics. Our main result is that when opinions represent action choices, the opinion associated with the action that is the fastest to execute spreads in the population. Moreover, the spread of the best choice happens even when only a minority is initially advocating for it. The key elements involved in this process are consensus building and positive feedback. A foraging task that involves collective transport is used to illustrate the potential of the proposed approach.

@incollection{MdeO:OpinionDynamicsANTS10,
Author = {Marco A. {Montes~de~Oca} and Eliseo Ferrante and Nithin Mathews and Mauro Birattari and Marco Dorigo},
Booktitle = {LNCS 6234. Proceedings of the Seventh International Conference on Swarm Intelligence (ANTS 2010)},
Editor = {M. Dorigo and others},
Title = {Opinion Dynamics for Decentralized Decision-Making in a Robot Swarm},
Pages = {252--263},
Address={Berlin, Germany},
Publisher={Springer},
Year = {2010}
}

To design high performing algorithms in swarm intelligence requires frequently the choice of appropriate parameter settings. Frequently, the parameters to be set are either continuous ones or have a large integer domain. For such tasks it seems therefore interesting to apply state-of-the-art continuous optimization algorithms instead of using a tedious and error-prone hands-on approach. In this article we study the performance of various continuous optimization algorithms for this algorithm configuration task using various case studies of algorithms to be configured from the swarm intelligence literature.

@incollection{Yuan:ParameterTuningANTS10,
Author = {Zhi Yuan and Marco A. {Montes~de~Oca} and Thomas St{\"u}tzle and Mauro Birattari},
Booktitle = {LNCS 6234. Proceedings of the Seventh International Conference on Swarm Intelligence (ANTS 2010)},
Editor = {M. Dorigo and others},
Title = {Modern Continuous Optimization Algorithms for Tuning Real and Integer Algorithm Parameters},
Pages = {204--215},
Address={Berlin, Germany},
Publisher={Springer},
Year = {2010}
}

In nature, there are examples of large groups of animals that are capable of making optimal collective-level decisions without the need for global control or information. Understanding the underlying mechanisms of such decentralized decision-making processes may help us to design artificial systems that exhibit some of the desirable properties, like scalability or fault tolerance, that are usually observed in these natural systems. In this paper, we show how a simple social influence mechanism, based on the binary particle swarm optimization algorithm, can make a whole population of agents achieve consensus on one of two possible choices in a completely decentralized way. Furthermore, we show that, if the conditions for achieving consensus are met and each choice is bound to an action that takes time to perform, the population converges to the choice associated with the shortest execution time. We illustrate the applicability of the decision-making mechanism presented in this paper on an example scenario in swarm robotics.

@incollection{MdeO:ECAL09Workshop,
Author = {Marco A. {Montes de Oca} and Eliseo Ferrante and Nithin Mathews and Mauro Birattari and Marco Dorigo},
Booktitle = {Proceedings of the Workshop on Organisation, Cooperation and Emergence in Social Learning Agents of the European Conference on Artificial Life (ECAL 2009)},
Editor = {D. Curran and C. O'Riordan},
Title = {Optimal Collective Decision-Making through Social Influence and Different Action Execution Times}
}

Particle swarm optimization (PSO) algorithms are used for solving real parameter optimization problems that may be characterized by discontinuities, plateaus, multiple local optima, and other features. In most PSO variants, the swarm is homogeneous, that is, all particles use the same search mechanism; however, given the possible complex nature of the solution space, a single search mechanism can be more effective in certain areas of the search space than in others. Thus, particle homogeneity can negatively affect the performance of PSO algorithms. A possible way to tackle this issue, is to use heterogeneous PSO algorithms, which are composed of particles that use different search mechanisms. In this paper, we propose a number of different strategies to allow particles to select, from a set of two candidates, their search mechanism according to information obtained during the optimization process. We show preliminary results and outline future research.

@incollection{Spenevello:SLSDS09,
Author = {Paolo Spanevello and Marco A. {Montes de Oca}},
Booktitle = {Proceedings of SLS-DS 2009: Doctoral Symposium on Engineering Stochastic Local Search Algorithms},
Editor = {F. Hutter and M. A. {Montes de Oca}},
Title = {Experiments on Adaptive Heterogeneous PSO Algorithms},
Pages={36--40},
Publisher = {IRIDIA, Universit{\'e} Libre de Bruxelles},
Address = {Brussels, Belgium},
Note={Technical Report No. TR/IRIDIA/2009-024}
}

Inferring gene regulatory networks from expression profiles is a challenging problem that has been tackled using many different approaches. When posed as an optimization problem, the goal is often to minimize the value of an error measure, usually the mean squared error or the relative squared error, between the real profiles and those generated with a model whose parameters are the variables being optimized. In this paper, we use recurrent neural networks to model regulatory interactions and study systematically the ``fitness landscape'' that results from measuring the relative squared error between the target and the inferred gene expression profiles. Although the results of the study indicate that the generated landscapes have a positive fitness-distance correlation, the error values span several orders of magnitude over very short distance variations. This suggests that the fitness landscape features extremely deep valleys, which can make general-purpose state-of-the-art continuous optimization algorithms exhibit a very poor performance. Further results obtained from an analysis based on perturbations of different magnitude to the optimal network topology, support approaches in which the spaces of network topologies and of network parameters are decoupled.

@incollection{SamboGRN:EA,
Author = {Francesco Sambo and Marco A. {Montes de Oca} and Barbara {Di Camillo} and Thomas St{\"u}tzle},
Booktitle = {LNCS 5975. Proceedings of the 9th International Conference on Artificial Evolution (EA 2009)},
Editor = {P. Collet and others},
Title = {On the Difficulty of Inferring Gene Regulatory Networks: A Study of the Fitness Landscape Generated by Relative Squared Error},
Pages = {74--85},
Address={Berlin,Germany},
Publisher={Springer},
Year = {2009}
}

Particle swarm optimization (PSO) is a swarm intelligence technique originally inspired by models of flocking and of social influence that assumed homogeneous individuals. During its evolution to become a practical optimization tool, some heterogeneous variants have been proposed. However, heterogeneity in PSO algorithms has never been explicitly studied and some of its potential effects have therefore been overlooked. In this paper, we identify some of the most relevant types of heterogeneity that can be ascribed to particle swarms. A number of particle swarms are classified according to the type of heterogeneity they exhibit, which allows us to identify some gaps in current knowledge about heterogeneity in PSO algorithms. Motivated by these observations, we carry out an experimental study of a couple of heterogeneous particle swarms each of which is composed of two kinds of particles. Directions for future developments on heterogeneous particle swarms are outlined.

@incollection{MonPenStu-etal2009:cec,
Address = {Piscataway, NJ},
Author = {Marco A. {Montes de Oca} and Jorge Pe{\~n}a and Thomas St{\"u}tzle and Carlo Pinciroli and Marco Dorigo},
Booktitle = {IEEE Congress on Evolutionary Computation -- CEC 2009},
Editor = {P. Haddow and others},
Pages = {698--705},
Publisher = {IEEE Press},
Title = {Heterogeneous Particle Swarm Optimizers},
Year = {2009}}

We present an algorithm that is inspired by theoretical and empirical results in social learning and swarm intelligence research. The algorithm is based on a framework that we call incremental social learning. In practical terms, the algorithm is a hybrid between a local search procedure and a particle swarm optimization algorithm with growing population size. The local search procedure provides rapid convergence to good solutions while the particle swarm algorithm enables a comprehensive exploration of the search space. We provide experimental evidence that shows that the algorithm can find good solutions very rapidly without compromising its global search capabilities.

@incollection{IPSOLSMdeO08,
author = "Marco A. {Montes~de~Oca} and Ken {Van~den~Enden} and Thomas St{\"u}tzle",
title = "Incremental Particle Swarm-Guided Local Search for Continuous Optimization",
booktitle = "LNCS 5296. Proceedings of the International Workshop on Hybrid Metaheuristics. HM 2008",
editor = "M. J. Blesa and others",
pages = "72--86",
address=" Berlin,Germany",
publisher="Springer",
year = "2008"
}

Social learning is a mechanism that allows individuals to acquire knowledge from others without incurring the costs of acquiring it individually. Individuals that learn socially can thus spend their time and energy exploiting their knowledge or learning new things. In this paper, we adapt these ideas for their application to both optimization and multiagent learning. The approach consists of a growing population of agents that learn socially as they become part of the main population. We find that learning socially in an incremental way can speed up the optimization and learning processes, as well as improve the quality of the solutions and strategies found.

@inproceedings{ISLMdeO08,
author = "Marco A. {Montes~de~Oca} and Thomas St{\"u}tzle",
title = "Towards Incremental Social Learning in Optimization and Multiagent Systems",
booktitle = "Proceedings of the Evolutionary Computation and Multi-Agent Systems and Simulation (ECoMASS) Workshop of the Genetic and Evolutionary Computation Conference (GECCO 2008)",
editor = "W. Rand, S. G. Ficici, and R. Riolo",
pages = "1939--1944",
publisher="ACM",
year = "2008"
}

The fully informed particle swarm optimization algorithm (FIPS) is very sensitive to changes in the population topology. The velocity update rule used in FIPS considers all the neighbors of a particle to update its velocity instead of just the best one as it is done in most variants. It has been argued that this rule induces a random behavior of the particle swarm when a fully connected topology is used. This argument could explain the often observed poor performance of the algorithm under that circumstance.

In this paper we study experimentally the convergence behavior of the particles in FIPS when using topologies with different levels of con- nectivity. We show that the particles tend to search a region whose size decreases as the connectivity of the population topology increases. We therefore put forward the idea that spatial convergence, and not a ran- dom behavior, is the cause of the poor performance of FIPS with a fully connected topology. The practical implications of this result are explored.

@inproceedings{FIPSMdeO08,
author = "Marco A. {Montes~de~Oca} and Thomas St{\"u}tzle",
title = "Convergence Behavior of the Fully informed Particle Swarm Optimization Algorithm",
booktitle = "Proceedings of the Genetic and Evolutionary Computation Conference (GECCO 2008)",
editor = "M. Keijzer and others",
pages = "71--78",
publisher="ACM",
year = "2008"
}

Many automatically-synthesized programs have, like their hand-made counterparts, numerical parameters that need to be set properly before they can show an acceptable performance. Hence, any approach to the automatic synthesis of programs needs the ability to tune numerical parameters efficiently.

Grammatical Evolution (GE) is a promising grammar-based genetic programming technique that synthesizes numbers by concatenating digits. In this paper, we show that a naive application of this approach can lead to a serious number length bias that in turn affects efficiency. The root of the problem is the way the context-free grammar used by GE is defined. A simple, yet effective, solution to this problem is proposed.

@inproceedings{GEMdeO08,
author = "Marco A. {Montes~de~Oca}",
title = "Exposing a Bias Toward Short-Length Numbers in Grammatical Evolution",
booktitle = "LNCS 4971. Proceedings of the Eleventh European Conference on Genetic Programming. EuroGP 2008",
editor = "M. O'Neill, L. Vanneschi, S. Gustafson, A. I. Esparcia-Alcázar, I. De Falco, A. Della Cioppa, and E. Tarantino",
pages = "278--288",
address=" Berlin",
publisher="Springer",
year = "2008"
}

Particle Swarm Optimization (PSO) is primarily used for solving continuous optimization problems. Over the years and as a result of the interest on PSO, many different variants of the original algorithm have been proposed and, more often than not, it is claimed that the new variant is superior in some way. It follows that if two variants differ only by some algorithmic components, then their difference in performance can be ascribed to differences in these components. A question then arises: Can we compose a high-performing PSO variant using components present in other variants? In this paper, we present an attempt to answer this question. The performance of the resulting composite variant suggests that the answer is positive. We believe that algorithm composition can help in devising effective optimization algorithms.

@incollection{MdO:SLSDS07,
Author = {Marco A. {Montes de Oca} and Thomas St{\"u}tzle and Mauro Birattari and Marco Dorigo},
Booktitle = {Proceedings of the Doctoral Symposium on Engineering Stochastic Local Search Algorithms -- SLS-DS 2007},
Editor = {E. Ridge and T. St{\"u}tzle and M. Birattari and H. H. Hoos},
Title = {Composing Particle Swarm Optimization Algorithms},
Pages = {6--10},
Publisher = {IRIDIA, Universit{\'e} Libre de Bruxelles},
Address = {Brussels, Belgium},
Note={Technical Report No. TR/IRIDIA/2007-014}
}

In this paper we present an estimation of distribution particle swarm optimization algorithm that borrows ideas from recent developments in ant colony optimization which can be considered an estimation of distribution algorithm. In the classical particle swarm optimization algorithm, particles exploit their individual memory to explore the search space. However, the swarm as a whole has no means to exploit its collective memory (represented by the array of previous best positions or pbests) to guide its search. This causes a re-exploration of already known bad regions of the search space, wasting costly function evaluations. In our approach, we use the swarm's collective memory to probabilistically guide the particles' movement towards the estimated promising regions in the search space. Our experiments show that this approach is able to find similar or better solutions than the canonical particle swarm optimizer with fewer function evaluations.

@inproceedings{EDAPSOMdeO06,
author = "Mudassar Iqbal and Marco A. {Montes~de~Oca}",
title = "An Estimation of Distribution Particle Swarm Optimization Algorithm",
booktitle = "LNCS 4150. Proceedings of the Fifth International Workshop on Ant Colony Optimization and Swarm Intelligence. ANTS 2006",
editor = "Marco Dorigo and Luca M. Gambardella and Mauro Birattari and Alcherio Martinoli and Riccardo Poli and Thomas St{\"u}tzle",
pages = "72--83",
address=" Berlin",
publisher="Springer",
year = "2006"
}

Version: 1.0
Author: Mudassar Iqbal and Marco A. Montes de Oca
Copyright (c) Mudassar Iqbal and Marco A. Montes de Oca, 2009
Download

In this paper we report an empirical comparison of some of the most influential Particle Swarm Optimization (PSO) algorithms based on run-length distributions (RLDs). The advantage of our approach over the usual report pattern (average iterations to reach a predefined goal, success rates, and standard deviations) found in the current PSO literature is that it is possible to evaluate the performance of an algorithm on different application scenarios at the same time. The RLDs reported in this paper show some of the strengths and weaknesses of the studied algorithms and suggest ways of improving their performance.

@inproceedings{PSORLDMdeO06,
author = "Marco A. {Montes~de~Oca} and Thomas St{\"u}tzle and Mauro Birattari and Marco Dorigo",
title = "A Comparison of Particle Swarm Optimization Algorithms Based on Run-Length Distributions",
booktitle = "LNCS 4150. Proceedings of the Fifth International Workshop on Ant Colony Optimization and Swarm Intelligence. ANTS 2006",
editor = "Marco Dorigo and Luca M. Gambardella and Mauro Birattari and Alcherio Martinoli and Riccardo Poli and Thomas St{\"u}tzle",
pages = "1--12",
address=" Berlin",
publisher="Springer",
year = "2006"
}

Communication among agents in swarm intelligent systems and more generally in multiagent systems, is crucial in order to coordinate agents' activities so that a particular goal at the collective level is achieved. From an agent's perspective, the problem consists in establishing communication policies that determine what, when, and how to communicate with others. In general, communication policies will depend on the nature of the problem being solved. This means that the solvability of problems by swarm intelligent systems depends, among other things, on the agents' communication policies, and setting an incorrect set of policies into the agents may result in finding poor solutions or even in the unsolvability of problems. As a case study, this paper focus on the effects of letting agents use different communication policies in ant-based clustering algorithms. Our results show the effects of using different communication policies on the final outcome of these algorithms.

@inproceedings{EffectsMdeO05,
author = "Marco A. {Montes~de~Oca} and Leonardo Garrido and Jos\'e L. Aguirre",
title = "Effects of Inter-agent Communication in Ant-based Clustering Algorithms: A Case Study on Communication Policies in Swarm Systems",
booktitle = "LNAI 3789. Proceedings of the Fourth Mexican International Conference on Artificial Intelligence. MICAI 2005",
editor = "Alexander Gelbukh and Alvaro {de~Albornoz} and Hugo Terashima",
pages = "254--263",
address=" Berlin",
publisher="Springer",
year = "2005"
}

Conventional ant-based clustering algorithms and growing neural gas networks are combined to produce an unsupervised classification algorithm that exploits the strengths of both techiques. The ant-based clustering algorithm detects existing classes on a training data set, and at the same time, trains several growing neural gas networks. On a second stage, these networks are used to classify previously unseen input vectors into the classes detected by the ant-based algorithm. The proposed algorithm eliminates the need of changing the number of agents and the dimensions of the environment when dealing with large databases.

@inproceedings{HybridizationMdeO05,
author = "Marco A. {Montes~de~Oca} and Leonardo Garrido and Jos\'e L. Aguirre",
title = "An hybridization of an ant-based clustering algorithm with growing neural gas networks for classification tasks",
booktitle = "Proceedings of the 20th Annual ACM Symposium on Applied Computing. SAC 2005",
editor = {Hisham Haddad and Lorie M. Liebrock and Andrea Omicini and Roger L. Wainwright},
pages = "9-13",
publisher="ACM",
year = "2005"
}

This paper presents a first approach to study direct information exchange between agents in ant-based clustering by comparing the clustering process development generated by agents that maintain a short-term memory against the one generated by agents that share their memory with their peers whenever an encounter occurs on the environment. The information exchange can allow an agent to "change its mind" regarding the most favorable dropping location on the grid, based on the knowledge of another agent.

Our experimental evidence shows that this simple information exchange strategy improves the quality of the resultant clustering. This holds true, however, only for a small number of agents. This suggests that there is a critical number of agents that can exchange information, that when surpassed, the effects could even be detrimental.

@inproceedings{FirstApproachMdeO04,
author = "Marco A. {Montes~de~Oca} and Leonardo Garrido and Jos\'e L. Aguirre",
title = "A first approach to study the effects of direct information exchange between agents in ant-based clustering",
booktitle = "Proceedings of the First World Congress on Lateral Computing. WCLC 2004",
editor = {Suthikshn Kumar and Ajith Abraham and Jens Harnisch and Antony Satyadas},
publisher="World Federation on Lateral-Computing",
note = "Proceedings published in CDROM",
year = "2004"
}

Ant-based clustering algorithms are distributed self-organized knowledge discovery tools inspired by the collective behavior of insect colonies. They are based in stigmergic communication among participating agents, that is, on the nonintentional indirect influence on the agents behavior through local modifications of the environment. Inspired by the oral thophalaxis phenomenon observed in some ant species, two different knowledge exchange strategies between agents in ant-based clustering algorithms are investigated. The impact on the final clustering quality is evaluated by comparing the clustering process development generated by each strategy. The investigated knowledge exchange strategies are: simple local memory sharing and the shared use of environment maps. It is shown that benefits on the final clustering are directly related to the usefulness of the exchanged knowledge and on the number of participating agents.

@inproceedings{EffectsMdeO04,
author = "Marco A. {Montes~de~Oca} and Leonardo Garrido and Jos\'e L. Aguirre",
title = "On the effects of direct communication between agents in ant-based clustering algorithms",
booktitle = "Proceedings of the Fifth Iberoamerican Workshop on Multi-Agent Systems. Iberagents 2004",
notes = "Proceedings delivered only to workshop participants",
year = "2004"
}

A swarm intelligence system is a type of multiagent system with the following distinctive characteristics: (i) it is composed of a large number of agents, (ii) the agents that comprise the system are simple with respect to the complexity of the task the system is required to perform, (iii) its control relies on principles of decentralization and self-organization, and (iv) its constituent agents interact locally with one another and with their environment.

Interactions among agents, either direct or indirect through the environment in which they act, are fundamental for swarm intelligence to exist; however, there is a class of interactions, referred to as interference, that actually blocks or hinders the agents' goal-seeking behavior. For example, competition for space may reduce the mobility of robots in a swarm robotics system, or misleading information may spread through the system in a particle swarm optimization algorithm. One of the most visible effects of interference in a swarm intelligence system is the reduction of its efficiency. In other words, interference increases the time required by the system to reach a desired state. Thus, interference is a fundamental problem which negatively affects the viability of the swarm intelligence approach for solving important, practical problems.

We propose a framework called incremental social learning (ISL) as a solution to the aforementioned problem. It consists of two elements: (i) a growing population of agents, and (ii) a social learning mechanism. Initially, a system under the control of ISL consists of a small population of agents. These agents interact with one another and with their environment for some time before new agents are added to the system according to a predefined schedule. When a new agent is about to be added, it learns socially from a subset of the agents that have been part of the system for some time, and that, as a consequence, may have gathered useful information. The implementation of the social learning mechanism is application-dependent, but the goal is to transfer knowledge from a set of experienced agents that are already in the environment to the newly added agent. The process continues until one of the following criteria is met: (i) the maximum number of agents is reached, (ii) the assigned task is finished, or (iii) the system performs as desired. Starting with a small number of agents reduces interference because it reduces the number of interactions within the system, and thus, fast progress toward the desired state may be achieved. By learning socially, newly added agents acquire knowledge about their environment without incurring the costs of acquiring that knowledge individually. As a result, ISL can make a swarm intelligence system reach a desired state more rapidly.

We have successfully applied ISL to two very different swarm intelligence systems. We applied ISL to particle swarm optimization algorithms. The results of this study demonstrate that ISL substantially improves the performance of these kinds of algorithms. In fact, two of the resulting algorithms are competitive with state-of-the-art algorithms in the field. The second system to which we applied ISL exploits a collective decision-making mechanism based on an opinion formation model. This mechanism is also one of the original contributions presented in this dissertation. A swarm robotics system under the control of the proposed mechanism allows robots to choose from a set of two actions the action that is fastest to execute. In this case, when only a small proportion of the swarm is able to concurrently execute the alternative actions, ISL substantially improves the system's performance.

@phdthesis{Mon11,
author = {Marco A. {Montes de Oca}},
title = {Incremental Social Learning in Swarm Intelligence Systems},
school = {Universit{\'e} Libre de Bruxelles},
address = {Brussels, Belgium},
month = {July},
year = {2011}
}

Since the introduction of the first Particle Swarm Optimization algorithm by James Kennedy and Russell Eberhart in the mid-90's, many variants of the original algorithm have been proposed. However, there is no general agreement on which variant(s) could be considered the state-of-the-art in the field. This is, in part, due to a general lack of cross-comparisons among variants in the literature.

The work reported in this document was carried out with the goal of identifying the best-performing particle swarm optimization algorithms. For practical reasons, we could not compare all the available algorithmic variants. Instead, we focused on those that have been the most widely used variants. We also considered algorithms that incorporate some of the latest developments in field. The comparison of the chosen particle swarm optimization algorithms was based on run-length and solution-quality distributions. These analytical tools allow the comparison of stochastic optimization algorithms in terms of the probability of finding a solution to an optimization problem within some minimum acceptable solution quality and time bounds. This kind of analysis is helpful for measuring the solution-quality vs. computational effort tradeoff each algorithm presents.

Particle swarm optimizers are population-based optimization algorithms that rely on a population structure referred to as topology. We conduct an analysis of the effects that different population topologies (the most commonly used ones) have in their performance. We show how some topologies are better suited for application scenarios in which only short runs are allowed or affordable, and how other topologies are better suited for situations in which the only important factor is solution quality.

Run-length and solution-quality distributions also serve as diagnostic tools that help in the design of improved algorithms. After analyzing the behavior of some variants with these tools, we found opportunities for improvement and the resultant variants are presented and discussed.

@mastersthesis{PSOPerfMdeO06,
author = "Marco A. {Montes~de~Oca}",
title = "On the Performance of Particle Swarm Optimizers",
school = "Universit\'e Libre de Bruxelles",
address = "Brussels, Belgium",
month = "September",
note = "Dipl\^ome d'\'Etudes Approfondies en Sciences Appliqu\'ees",
year = "2006"
}

Ant-based clustering algorithms are knowledge discovery tools inspired by the collective behavior of social insect colonies. In these algorithms, insects are modeled as software agents that communicate with each other indirectly through the environment. This particular kind of communication is known as stigmergic communication.

In the classic ant-based clustering algorithm, a group of agents that exhibit the same behavior move randomly over a toroidal square grid. In the environment there are data objects that were initially scattered in a random fashion. The objects can be picked up, moved or dropped in any free location on the grid. An object is picked up with high probability if it is not surrounded by similar objects and is dropped with high probability if an agent's neighborhood is densely populated by other similar objects and its location is free. Here, stigmergy occurs when an object is placed next to another. The resultant structure is much more attractive to agents to drop other similar objects nearby.

However, stigmergy is not the only way social insects interact with each other. In most species, trophallaxis or liquid food exchange among members of the same colony, plays a key role in their social organization. Consider the case of some termite species which require intestinal protozoa to derive benefits from cellulose. Their early instar nymphs are fed either by oral or anal trophallaxis. The latter infects them with symbiotic protozoa or bacteria contained in the proctodeal liquid. The subsocial association result of this codependence has evolved into a complex social and morphological structure.

Inspired by the trophallaxis phenomenon observed in some ant and termite species, two different communication strategies among agents in ant-based clustering algorithms are investigated: (i) direct and (ii) indirect communication. The impact on the final clustering quality is evaluated by comparing the development of the clustering process generated by each strategy. It is shown that benefits on the final clustering are directly related to the usefulness of the exchanged information, its use, and on the number of participating agents.

@mastersthesis{EffectsMscMdeO05,
author = "Marco A. {Montes~de~Oca}",
title = "Effects on clustering quality of direct and indirect communication among agents in ant-based clustering algorithms",
school = "Instituto Tecnol\'ogico y de Estudios Superiores de Monterrey. Campus Monterrey.",
address = "Monterrey, N.L. M\'exico",
month = "May",
year = "2005"
}

In this paper, we present a continuous-time binary consensus protocol whereby entities connected via a directed ring topology solve the one-dimensional density classification problem. In our model, entities behave as non-ideal relays, that is, they have memory of the trajectory of an internal state variable, which gives them hysteretic properties. We show that this feature is necessary for the system to reach consensus on the state shared by the initial majority.

Social interactions underpin collective decision making in all animal societies. However, the actual mechanisms that animals use to achieve a collective decision differ among species. For example, ants use pheromones to bias the decisions of other ants; birds observe and match the velocity of their neighbors; humans match the speed of other people while driving (even above speed limits). Despite the differences among these (and other) collective decision-making mechanisms, some basic principles underlying them exist, like the tendency to conform to the actions or opinions of others. In our examples, by following pheromone trails ants effectively follow on their nestmates' steps, birds in a flock are more likely to fly in the same direction, and we humans, while we do not always agree, we do not like to be in permanent conflict with others and eventually seek ways to compromise. Therefore, this principle's basic operating mechanism is that an individual who is exposed to the actions or opinions of others tends to perform the actions, or have the same opinions of the observed individuals.

In this communication, I describe two basic collective decision-making mechanisms based on the principle outlined above. These mechanisms are tested in a setting that simulates a robotics scenario in which a group of robots must collectively find the shorter of two alternative paths between two areas without measuring travel times or distances. First, I describe a mechanism that consists of robots forming teams of three robots (or a greater odd-number of robots) that decide which path to use by locally using the majority rule. Then, I describe a mechanism that consists in individual robots increasing the tendency to choose either path based on the path recently taken by another robot. In both cases, the group collectively chooses the shorter path with high probability. These mechanisms have been proposed as swarm intelligence mechanisms for optimal collective decision-making.

@incollection{Mon13b,
Author = {Marco A. {Montes~de~Oca}},
Booktitle = {Proceedings of the 1st Multidisciplinary Conference on Reinforcement Learning and Decision Making},
Title = {Social Reinforcement for Collective Decision-Making Over Time},
Year = {2013}
}