GECCO logo

11th Workshop on Evolutionary Computation for the Automated Design of Algorithms (ECADA)

“It may be turtles all the way down, but the turtles get smaller.” -- Anonymous @ ECADA 2017

July 10-14, 2021 @ GECCO 2021

Note that GECCO 2021 will be an electronic-only conference due to COVID-19

Important Dates

April 12, 2021 Workshop paper submission deadline
April 26, 2021 Notification of acceptance
May 3, 2021 Camera-ready deadline
May 3, 2021 Author registration deadline

Related Events

Invited ECADA Talk

[Picture] Speaker: Gisele L. Pappa

Title: Understanding the Search Space of Automated Machine Learning Algorithms

Abstract: Automated Machine Learning (Auto-ML) is the cousin of Hyperheuristics for Machine Learning. It has become widely popular since the term was coined in 2013, when it was first used to build complete machine learning pipelines - a sequence of steps to solve a particular problem that may include both preprocessing (e.g., feature selection), classification, and postprocessing (e.g., ensemble-like methods). In the past years, the area has turned from searching ML pipelines to searching the architecture of complex neural networks, a field known as Neural Architecture Search (NAS). In both cases, the most popular search methods are mainly based on Bayesian Optimization or Evolutionary Algorithms, while reinforcement learning is also popular for NAS. However, the search space of AutoML problems, in general, is complex, including categorical, discrete, continuous, and conditional variables. This talk presents work that has been done to better understand these search spaces, looking mainly at how to define neighborhoods and generate measures of fitness correlation and neutrality. This is essential to grasp which methods are more promising in different scenarios and develop more appropriate search mechanisms to take advantage of the structure of these spaces.

Short Bio: Gisele Pappa is an Associate Professor in the Computer Science Department at UFMG, Brazil. She has served as a GECCO Self-* track co-chair in past editions and has also been responsible for both the tutorials and workshops at GECCO. She is an associate editor of Genetic Programming and Evolvable Machines journal and has an extensive publication record in the intersection of the machine learning and evolutionary computation areas. She has also been actively researching the use of EAs for automated machine learning (AutoML), and currently looks at the search spaces of these algorithms and how they can be effectively explored. Other research interests are in genetic programming and its applications to both classification and regression tasks focusing on applications for health data and also fraud detection.

Workshop Schedule

Accepted papers: Detailed Schedule To Be Announced

Description

The main objective of this workshop is to discuss hyper-heuristics and algorithm configuration methods for the automated generation and improvement of algorithms, with the goal of producing solutions (algorithms) that are applicable to multiple instances of a problem domain. The areas of application of these methods include optimization, data mining and machine learning. [1-18,23].

Automatically generating and improving algorithms by means of other algorithms has been the goal of several research fields, including Artificial Intelligence in the early 1950s, Genetic Programming since the early 1990s, and more recently automated algorithm configuration [1] and hyper-heuristics [2]. The term hyper-heuristics generally describes meta-heuristics applied to a space of algorithms. While Genetic Programming has most famously been used to this end, other evolutionary algorithms and meta-heuristics have successfully been used to automatically design novel (components of) algorithms. Automated algorithm configuration grew from the necessity of tuning the parameter settings of meta-heuristics and it has produced several powerful (hyper-heuristic) methods capable of designing new algorithms by either selecting components from a flexible algorithmic framework [3,4] or recombining them following a grammar description [5][9].

Although most evolutionary algorithms are designed to generate specific solutions to a given instance of a problem, one of the defining goals of hyper-heuristics is to produce solutions that solve more generic problems. For instance, while there are many examples of evolutionary algorithms for evolving classification models in data mining and machine learning, a genetic programming hyper-heuristic has been employed to create a generic classification algorithm which in turn generates a specific classification model for any given classification dataset, in any given application domain [8]. In other words, the hyper-heuristic is operating at a higher level of abstraction compared to how most search methodologies are currently employed; i.e., it is searching the space of algorithms as opposed to directly searching in the problem solution space [9], raising the level of generality of the solutions produced by the hyper-heuristic evolutionary algorithm. In contrast to standard Genetic Programming, which attempts to build programs from scratch from a typically small set of atomic functions, generative hyper-heuristic methods specify an appropriate set of primitives (e.g., algorithmic components) and allow evolution to combine them in novel ways as appropriate for the targeted problem class. While this allows searches in constrained search spaces based on problem knowledge, it does not in any way limit the generality of this approach as the primitive set can be selected to be Turing-complete. Typically, however, the initial algorithmic primitive set is composed of primitive components of existing high-performing algorithms for the problems being targeted; this more targeted approach very significantly reduces the initial search space, resulting in a practical approach rather than a mere theoretical curiosity. Iterative refining of the primitives allows for gradual and directed enlarging of the search space until convergence.

As meta-heuristics are themselves a type of algorithm, they too can be automatically designed employing hyper-heuristics. For instance, in 2007, Genetic Programming was used to evolve mate selection in evolutionary algorithms [11]; in 2011, Linear Genetic Programming was used to evolve crossover operators [12]; more recently, Genetic Programming was used to evolve complete black-box search algorithms [13,14,16], SAT solvers [22], and FuzzyART category functions [23]. Moreover, hyper-heuristics may be applied before deploying an algorithm (offline) [5] or while problems are being solved (online) [9], or even continuously learn by solving new problems (life-long) [19]. Offline and life-long hyper-heuristics are particularly useful for real-world problem solving where one can afford a large amount of a priori computational time to subsequently solve many problem instances drawn from a specified problem domain, thus amortizing the a priori computational time over repeated problem solving. Recently, the design of Multi-Objective Evolutionary Algorithm components was automated [21].

Very little is known yet about the foundations of hyper-heuristics, such as the impact of the meta-heuristic exploring algorithm space on the performance of the thus automatically designed algorithm. An initial study compared the performance of algorithms generated by hyper-heuristics powered by five major types of Genetic Programming [18]. Another avenue for research is investigating the potential performance improvements obtained through the use of asynchronous parallel evolution to exploit the typical large variation in fitness evaluation times when executing hyper-heuristics [20].

[1]
Edmund K. Burke, Michel Gendreau, Matthew Hyde, Graham Kendall, Gabriela Ochoa, Ender Özcan, & Rong Qu. (2013). Hyper-heuristics: A survey of the state of the art. Journal of the Operational Research Society, 64(12), 1695-1724.
[2]
Holger H. Hoos (2012). Automated algorithm configuration and parameter tuning. In Autonomous search (pp. 37-71). Springer Berlin Heidelberg. doi:10.1007/978-3-642-21434-9_3.
[3]
KhudaBukhsh, A. R., Xu, L., Holger H. Hoos & Leyton-Brown, K. (2009). SATenstein: Automatically Building Local Search SAT Solvers from Components. In IJCAI, 9, 517-524.
[4]
Manuel López-Ibáñez and Thomas Stützle. (2012). The automatic design of multiobjective ant colony optimization algorithms. IEEE Transactions on Evolutionary Computation, 16(6):861-875.
[5]
Mascia, F., Manuel López-Ibáñez, Dubois-Lacoste, J., & Thomas Stützle. (2014). Grammar-based generation of stochastic local search heuristics through automatic algorithm configuration tools. Computers & operations research, 51, 190-199.
[6]
William B. Langdon and Mark Harman. Genetically Improving 50000 Lines of C++. Research Note , RN/12/09, Department of Computer Science, University College London, Gower Street, London WC1E 6BT, UK, 2012.
[7]
Justyna Petke, Mark Harman, William B. Langdon, and Westley Weimer. Using Genetic Improvement & Code Transplants to Specialise a C++ Program to a Problem Class Proceedings of the 17th European Conference on Genetic Programming, EuroGP 2014, Granada, Spain, 2014. Springer Verlag.
[8]
Gisele L. Pappa and Alex A. Freitas. Automating the Design of Data Mining Algorithms: An Evolutionary Computation Approach, Springer, Natural Computing Series, 2010.
[9]
Edmund K. Burke, Matthew Hyde, Graham Kendall and John Woodward. A genetic programming hyper-heuristic approach for evolving 2-D strip packing heuristics. In IEEE Transactions on Evolutionary Computation, 14(6):942-958, December 2010.
[10]
M. Oltean and D. Dumitrescu. Evolving TSP heuristics using multi expression programming. In: Computational Science - ICCS 2004, Lecture Notes in Computer Science 3037, pp. 670-673. Springer, 2004.
[11]
Ekaterina A. Smorodkina and Daniel R. Tauritz. Toward Automating EA Configuration: the Parent Selection Stage. In Proceedings of CEC 2007 - IEEE Congress on Evolutionary Computation, pages 63-70, Singapore, September 25-28, 2007.
[12]
Brian W. Goldman and Daniel R. Tauritz. Self-Configuring Crossover. In Proceedings of the 13th Annual Conference Companion on Genetic and Evolutionary Computation (GECCO '11), pages 575-582, Dublin, Ireland, July 12-16, 2011.
[13]
Matthew A. Martin and Daniel R. Tauritz. Evolving Black-Box Search Algorithms Employing Genetic Programming. In Proceedings of the 15th Annual Conference Companion on Genetic and Evolutionary Computation (GECCO '13), pages 1497-1504, Amsterdam, The Netherlands, July 6-10, 2013.
[14]
Matthew A. Martin and Daniel R. Tauritz. A Problem Configuration Study of the Robustness of a Black-Box Search Algorithm Hyper-Heuristic. In Proceedings of the 16th Annual Conference Companion on Genetic and Evolutionary Computation (GECCO '14), pages 1389-1396, Vancouver, BC, Canada, July 12-16, 2014.
[15]
John R. Woodward and Jerry Swan, "The automatic generation of mutation operators for genetic algorithms", in Proceedings of the 14th international conference on Genetic and evolutionary computation conference, 2012.
[16]
Matthew A. Martin and Daniel R. Tauritz. Hyper-Heuristics: A Study On Increasing Primitive-Space. In Proceedings of the 17th Annual Conference Companion on Genetic and Evolutionary Computation (GECCO '15), pages 1051-1058, Madrid, Spain, July 11-15, 2015.
[17]
Su Nguyen and Mengjie Zhang and Mark Johnston and Kay Chen Tan. Automatic Design of Scheduling Policies for Dynamic Multi-objective Job Shop Scheduling via Cooperative Coevolution Genetic Programming. IEEE Transactions on Evolutionary Computation, 18(2):193-208, April 2014.
[18]
Sean Harris, Travis Bueter, and Daniel R. Tauritz. A Comparison of Genetic Programming Variants for Hyper-Heuristics. In Proceedings of the 17th Annual Conference Companion on Genetic and Evolutionary Computation (GECCO '15), pages 1043-1050, Madrid, Spain, July 11-15, 2015.
[19]
Kevin Sim, Emma Hart, and Ben Paechter. A Lifelong Learning Hyper-Heuristic for Bin-Packing Evolutionary Computation, MIT Press Evolutionary Computation, 23(1):37-67, March 2015.
[20]
Alex R. Bertels, and Daniel R. Tauritz. Why Asynchronous Parallel Evolution is the Future of Hyper-heuristics: A CDCL SAT Solver Case Study. In Proceedings of the 18th Annual Conference Companion on Genetic and Evolutionary Computation (GECCO '16), pages 1359-1365, Denver, Colorado, U.S.A., July 20-24, 2016.
[21]
Leonardo C. T. Bezerra, Manuel López-Ibáñez, and Thomas Stützle. Automatic Component-Wise Design of Multi-Objective Evolutionary Algorithms. IEEE Transactions on Evolutionary Computation, 20(3):403-417, 2016.
[22]
Marketa Illetskova, Alex R. Bertels, Joshua M. Tuggle, Adam Harter, Samuel Richter, Daniel R. Tauritz, Samuel Mulder, Denis Bueno, Michelle Leger and William M. Siever. Improving Performance of CDCL SAT Solvers by Automated Design of Variable Selection Heuristics. Accepted for publication in the proceedings of the 2017 IEEE Symposium Series on Computational Intelligence (SSCI 2017), Honolulu, Hawaii, U.S.A., November 27 - December 1, 2017.
[23]
Islam Elnabarawy, Daniel R. Tauritz and Donald C. Wunsch. Evolutionary Computation for the Automated Design of Category Functions for Fuzzy ART: An Initial Exploration. In Proceedings of the 19th Annual Conference Companion on Genetic and Evolutionary Computation (GECCO '17), pages 1133-1140, Berlin, Germany, July 15-19, 2017.

Call for Papers

We welcome original submissions on all aspects of Evolutionary Computation for the Automated Design of Algorithms, in particular, evolutionary computation methods and other hyper-heuristics for the automated design, generation or improvement of algorithms that can be applied to any instance of a target problem domain. Relevant methods include methods that evolve whole algorithms given some initial components as well as methods that take an existing algorithm and improve it or adapt it to a specific domain. Another important aspect in automated algorithm design is the definition of the primitives that constitute the search space of hyper-heuristics. These primitives should capture the knowledge of human experts about useful algorithmic components (such as selection, mutation and recombination operators, local searches, etc) and, at the same time, allow the generation of new algorithm variants. Examples of the application of hyper-heuristics, including genetic programming and automatic configuration methods, to such frameworks of algorithmic components are of interest to this workshop, as well as the (possibly automatic) design of the algorithmic components themselves and the overall architecture of metaheuristics. Therefore, relevant topics include (but are not limited to):

Paper Submission

Workshop papers must be submitted using the GECCO submission site. After login, the authors need to select the "Workshop Paper" submission form. In the form, the authors must select the workshop they are submitting to. To see a sample of the "Workshop Paper" submission form, go to the GECCO submission site and select "Sample Submission Forms". Submitted papers must not exceed 8 pages (excluding references) and are required to be in compliance with the GECCO 2021 Papers Submission Instructions. It is recommended to use the same templates as the papers submitted to the main tracks. Each paper submitted to this workshop will be rigorously reviewed in a double-blind review process. In other words, authors should not know who the reviewers of their work are and reviewers should not know who the authors are. To this end, the following information is very important: Submitted papers should be ANONYMIZED. This means that they should NOT contain any element that may reveal the identity of their authors. This includes author names, affiliations, and acknowledgments. Moreover, any references to any of the author's own work should be made as if the work belonged to someone else. All accepted papers will be presented at the ECADA workshop and appear in the GECCO 2021 Conference Companion Proceedings. By submitting a paper, the author(s) agree that, if their paper is accepted, they will:

Program Committee

Co-Chairs (in alphabetical order)

[Picture]

E-mail: manuel.lopez-ibanez@uma.es

Dr. López-Ibáñez is Senior Distinguished Researcher at the University of Málaga (Spain) and a Senior Lecturer (Associate Professor) in the Decision and Cognitive Sciences Research Centre at the Alliance Manchester Business School, University of Manchester, UK. He received the M.S. degree in computer science from the University of Granada, Granada, Spain, in 2004, and the Ph.D. degree from Edinburgh Napier University, U.K., in 2009. He has published 27 journal papers, 9 book chapters and 48 papers in peer-reviewed proceedings of international conferences on diverse areas such as evolutionary algorithms, ant colony optimization, multi-objective optimization, pump scheduling and various combinatorial optimization problems. His current research interests are experimental analysis and automatic design of stochastic optimization algorithms, for single and multi-objective optimization. He is the lead developer and current maintainer of the irace software package (http://iridia.ulb.ac.be/irace).
[Picture]

E-mail: dtauritz@acm.org

Daniel R. Tauritz is an Associate Professor in the Department of Computer Science and Software Engineering at Auburn University (AU), Interim Director and Chief Cyber AI Strategist of the Auburn Cyber Research Center, the founding Head of AU's Biomimetic Artificial Intelligence Research Group (BioAI Group), a cyber consultant for Sandia National Laboratories, a Guest Scientist at Los Alamos National Laboratory (LANL), and founding academic director of the LANL/AU Cyber Security Sciences Institute (CSSI). He received his Ph.D. in 2002 from Leiden University. His research interests include the design of generative hyper-heuristics and self-configuring evolutionary algorithms and the application of computational intelligence techniques in cyber security, critical infrastructure protection, and program understanding. He was granted a US patent for an artificially intelligent rule-based system to assist teams in becoming more effective by improving the communication process between team members.

[Picture]

E-mail: j.woodward@qmul.ac.uk

John R. Woodward is a Lecturer in at the Queen Mary University of London. Formerly he was a lecturer at the University of Stirling, within the CHORDS group and was employed on the DAASE project. Before that he was a lecturer for four years at the University of Nottingham. He holds a BSc in Theoretical Physics, an MSc in Cognitive Science and a PhD in Computer Science, all from the University of Birmingham. His research interests include Automated Software Engineering, particularly Search Based Software Engineering, Artificial Intelligence/Machine Learning and in particular Genetic Programming. He has over 50 publications in Computer Science, Operations Research and Engineering which include both theoretical and empirical contributions, and given over 50 talks at international conferences and as an invited speaker at universities. He has worked in industrial, military, educational and academic settings, and been employed by EDS, CERN and RAF and three UK Universities.

Previous ECADA workshops