Ant Colony Optimization and Swarm Intelligence - ACO+SI

SPSO2011 - Analysis of stability, local convergence, and rotation sensitivity
Mohammad Reza Bonyadi, Zbigniew Michalewicz

Praise (by Marco A. Montes de Oca and Konstantinos E. Parsopoulos): This paper sheds light on parameter selection of an important PSO variant by studying its stability, convergence, and rotational invariance. Unlike other works on similar topics, the dimensionality of the problem is shown to play a significant role in parameter setting. We expect the community to greatly benefit from Bonyadi and Michalewicz's work.

Anticipatory Stigmergic Collision Avoidance under Noise
Friedrich Burkhard von der Osten, Michael Kirley, Tim Miller

Praise (by Marco A. Montes de Oca and Konstantinos E. Parsopoulos): This paper introduces a novel scheme based on stigmergic principles where pheromones are not related to previous actions but rather on intended future actions. Through this scheme it is possible to reduce conflicts among agents, and therefore, it may be used for increasing the efficiency of a swarm. The scheme is shown to be very robust in agent-based simulations of reactive path planning and obstacle avoidance under noise.

Artificial Immune Systems - AIS

On the Runtime Analysis of Stochastic Ageing Mechanisms
Pietro S. Oliveto, Dirk Sudholt

Praise (by Emma Hart and Christine Zarges): The paper considers the use of different ageing operators within artificial immune systems in both static and dynamic settings. It contributes significantly to the understanding of how and why ageing can be useful by presenting an elegant and sound theoretical analysis. The paper which is accessible to both, theoreticians and practitioners, is an excellent contribution that will highly benefit not only the field of artificial immune systems but the wider GECCO community by introducing new concepts to the algorithmic toolbox to tackle practical problems.

Artificial Life, Robotics, and Evolvable Hardware - ALIFE

Overcoming Deception in Evolution of Cognitive Behavior
Joel Lehman, Risto Miikkulainen

Praise (by Thomas Schmickl and Ken Stanley ): The paper was recognized by reviewers for revealing a potentially important insight into the conditions that lead to the evolution of cognitive-level behavior. In short, objective-driven or goal-oriented fitness functions may be less effective than more open-ended evolution for achieving cognitive capabilities. The reviewers’ comments stand for themselves: “A truly excellent paper.” “Really cool paper.” “A very good work on an important topic."

A Novel Human-Computer Collaboration: Combining Novelty Search with Interactive Evolution
Brian Woolley, Kenneth Stanley

Praise (by Thomas Schmickl ): This paper describes a novel method named "NA-IEC" which is combining interactive evolution and novelty search. This novel approach is shown to increase the speed of evolution compared to other state-of-the-art methods. The study contains several well-motivated experiments with a self-critical discussion of the gained results. This is leading to interesting conclusions and a strong argumentation of this novel method, bringing it into the context of the work of the scientific community.

Digital Entertainment and Arts - DETA

Evolving Multimodal Behavior With Modular Neural Networks in Ms. Pac-Man
Jacob Schrum, Risto Miikkulainen

Praise (by Christian Jacob and Julian Togelius): This paper convincingly shows not only that Ms. Pac-Man is a multimodal problem, but also that this can be solved with a modular neural network and that the task decomposition can be found automatically through evolution. These results are important both for the study of modular neural networks, and for the development of high-performing game and robot controllers.

Evolutionary Combinatorial Optimization and Metaheuristics - ECOM

A Heuristic Approach to Schedule Reoptimization in the Context of Interactive Optimization
David Meignan

Praise (by Günther Raidl and Thomas Stützle): This paper proposes a new heuristic approach to the re-optimization of schedules, which is a task that frequently arises in practical situations, where experts assess and adjust solutions before taking final decisions. A detailed experimental study clearly establishes the effectiveness and the adequacy of the proposed approach.

Revised Analysis of the (1+1) EA for the Minimum Spanning Tree Problem
Carsten Witt

Praise (by Günther Raidl and Thomas Stützle): Making use of recent, advanced proof techniques, the paper provides improved upper bounds for the running time of a 1+1 EA on the minimum spanning tree problem, a paradigmatic combinatorial problem optimization problem that has received significant attention in recent theoretical work in evolutionary combinatorial computation.

Evolutionary Machine Learning -EML

Salient Object Detection Using Learning Classifier Systems that Compute Action Mappings
Muhammad Iqbal, Syed Naqvi, Will Neil Browne, Christopher Hollitt, Mengjie Zhang

Praise (by Jaume Bacardit and Tom Schaul): This is a solid paper showing the applicability of evolutionary machine learning methods to real-world problems, particularly in the computer vision task of salient objects detection. The method presented in this paper leverages on existing computer vision methods and provides added-value on top of these in the form of white-box prediction models from which it is straightforward to extract explanations of the method’s predictions.

Evolutionary Multiobjective Optimization - EMO

Inverted PBI in MOEA/D and its Impact on the Search Performance on Multi and Many-Objective Optimization
Hiroyuki Sato

Praise (by Dimo Brockhoff and Joshua D. Knowles): The paper "Inverted PBI in MOEA/D and its Impact on the Search Performance on Multi- and Many-Objective Optimization" adds a simple-looking idea (the kind you wish you'd had yourself) to an already excellent algorithm, MOEA/D, to improve diversity preservation. The idea is to angle the contour lines of the utility functions so that solutions are strongly rewarded for staying close to a particular region in objective space, defined by the weight vectors. Results are promising on a range of problems, especially many-objective ones.

Evolution Strategies and Evolutionary Programming - ESEP

A Computationally Efficient Limited Memory CMA-ES for Large Scale Optimization
Ilya Loshchilov

Praise (by Anne Auger and Tobias Glasmachers):The paper opens a new direction for scaling covariance matrix adaptation to search and optimization in extremely high-dimensional search spaces. The algorithm allows for the application of modern evolution strategies to novel application areas and hence addresses a highly relevant problem. The presented linear time low-rank covariance matrix update is elegant, theoretically sound, and effective.

Genetic Algorithms - GA

A Fixed Budget Analysis of Randomized Search Heuristics for the Traveling Salesperson Problem
Samadhi Nallaperuma, Frank Neumann, Dirk Sudholt

Praise (by Kalyanmoy Deb and Thomas Jansen):The paper is the first to apply the perspective of fixed-budget computations to an NP-hard optimization problem. It tackles one of the best known and most studied such problems, the traveling salesperson problem (TSP), and proves a number of meaningful and relevant lower bounds for the performance of a simple evolutionary algorithm and local search for large classes of TSP instances. The paper is an excellent example of theoretical research delivering tangible results by not restricting the analysis to toy problems.

Parameter-less Population Pyramid
Brian W. Goldman, William F. Punch

Praise (by Kalyanmoy Deb and Thomas Jansen): The paper is a new approach on the old problem of designing a genetic algorithm that gets rid of the problem of finding appropriate parameter setting. The new algorithm shows very impressive performance on a range of different test problems and in comparison with a number of competing algorithms. Among the strengths of the paper are presenting the algorithm together with a thorough empirical evaluation and theoretical analysis.

Generative and Developmental Systems - GDS

Some distance measures for morphological diversification in Generative Evolutionary Robotics
Eivind Samuelsen, Kyrre Glette

Praise (by Michael Palmer and Sebastian Risi):Papers advocating diversity or novelty search (basing fitness of a solution, partially or entirely, on how different it is from previously visited solutions) have often applied it in low-dimensional spaces using simple metrics of phenotypic distance, but it has not been clear whether the benefits of rewarding diversity can be extended to high-dimensional problems. In contrast, this paper studies a complex problem domain - locomotion in physically simulated robots with both evolved morphology and control systems. The results highlight the care required to define a useful phenotypic distance metric in high-dimensional spaces: three apparently reasonable metrics hurt overall performance; one helps significantly.

Genetic Programming - GP

Evolving "Less-Myopic" Scheduling Rules for Dynamic Job Shop Scheduling with Genetic Programming
Rachel J. Hunt, Mark Johnston, Mengjie Zhang

Praise (by Malcolm Heywood and William Langdon): Dispatching rules are frequently used in job-shop scheduling (JSS) tasks as a local heuristic for deciding what task to perform next at each machine comprising a (job-shop) manufacturing facility. However, they also lack many properties found in more global scheduling mechanisms. This work addresses the question of evolving programs to provide less myopic (more globally general) dispatching rules that still retain the local operational properties that make dispatching rules popular from an application perspective.

Behavioral Programming: A Broader and More Detailed Take on Semantic GP
Krzysztof Krawiec, Una-May O'Reilly

Praise (by Malcolm Heywood and William Langdon): Generic behavioural fitness measures that guide evolution towards the overall objective without leading to local minima are potentially fundamental to scaling GP to larger tasks. An approach is proposed that uses lists of internal program state during execution to bias the identification of code segments for archiving and reuse. Evaluation over 35 benchmarks demonstrates significant improvements while little computational overhead is encountered.

Parallel Evolutionary Systems - PES

Design and Analysis of Adaptive Migration Intervals in Parallel Evolutionary Algorithms
Andrea Mambrini, Dirk Sudholt

Praise (by Stefano Cagnoni and Gabriel Luque): Theory is often guiltily neglected in most articles and research lines today, but there is still much to discuss and discover about parallel evolutionary systems. The authors of the paper we nominated propose a new migration scheme, a most relevant factor in the distributed parallel model, and theoretically analyze its influence on the behaviour of the parallel algorithm that relies on it. The method and the results described by the authors can definitely contribute to the creation of a more solid body of knowledge about parallel metaheuristics.

Real World Applications - RWA

Multi-Objective Routing Optimisation for Battery Powered Wireless Sensor Mesh Networks
Jonathan Edward Fieldsend, Richard Everson, Alma As-Aad, Mohammad Rahat

Praise (by Hitoshi Iba and Bernhard Sendhoff): The paper describes the adaptation and use of a multi-objective optimization algorithm to a real world application very well. The illustration of the MOO applied to WSNs for the Victoria & Albert Museum London is an excellent example for applying techniques from natural computation to realistic problem scenarios.

On Homogenization of Coal in Longitudinal Blending Beds
Pradyumn Kumar Shukla, Michael Cipold, Claus Bachmann, Hartmut Schmeck

Praise (by Hitoshi Iba and Bernhard Sendhoff): The paper is an exceptionally well written description of the adaptation of evolutionary computation to the needs of an industrially relevant real-world application. The problem description is comprehensible to the computational intelligence researcher, the algorithmic instantiation of the optimization method is well described and the analysis is scientifically sound.

Search Based Software Engineering - SBSE

On the Performance of Multiple Objective Evolutionary Algorithms for Software Architecture Discovery
Aurora Ramírez, José Raúl Romero, Sebastian Ventura

Praise (by Marouane Kessentini and Guenther Ruhe): This paper tackles the problem of finding the most suitable software architecture for a system, whose discovery is realized by means of evolutionary algorithms. The authors explore the performance (considered in terms of a number of different practical aspects) of a set of multi- and many-objective algorithms. The methodology is clearly described, and the empirical analysis for comparison between the five different algorithms is applied to the same experimental data using the same the encoding, genetic operators, and evaluation objectives.

THEORY

Evolution under Partial Information
Duc-Cuong Dang, Per Kristian Lehre

Praise (by Benjamin Doerr and Carsten Witt : The paper addresses, by mathematical means, a major challenge in evolutionary computation, namely how to efficiently solve problems when the fitness of individuals or populations can only partially be evaluated. Using advanced drift analysis and fitness level methods, the authors succeed in distilling conditions that allow efficient optimization despite such difficulties. Their results in particular indicate that large parent populations are beneficial.