ECADA 2012 - 2nd Workshop on Evolutionary Computation for the Automated Design of Algorithms

Although most of the evolutionary computation techniques are designed to generate specific solutions to a given instance of a problem, some of these techniques can be explored to solve more generic problems. For instance, while there are many examples of evolutionary algorithms for evolving classification models in data mining or machine learning, the work described in [1] used a genetic programming algorithm to create a generic classification algorithm which will, in turn, generate a specific classification model for any given classification dataset, in any given application domain.
Although the work in [1] consisted of evolving a complete data mining/machine learning algorithm, in the area of optimization this type of approach is named a hyper-heuristic. Hyper-heuristics are search methods that automatically select and combine simpler heuristics, creating a generic heuristic that is used to solve any instance of a given target type of optimization problem. Hence, hyper-heuristics search in the space of heuristics, instead of searching in the problem solution space [2], raising the level of generality of the solutions produced by the hyper-heuristic evolutionary algorithm. For instance, a hyper-heuristic can generate a generic heuristic for solving any instance of the traveling salesman problem, involving any number of cities and any set of distances associated with those cities [3]; whilst a conventional evolutionary algorithm would just evolve a solution to one particular instance of the traveling salesman problem, involving a predefined set of cities and associated distances between them.
Whether we name it an approach for automatically designing algorithms or hyper-heuristics, in both cases, a set of human designed procedural components or heuristics surveyed from the literature are chosen as a starting point (or as "building blocks") for the evolutionary search. Besides, new procedural components and heuristics can be automatically generated, depending on which components are first provided to the method.
The main objective of this workshop is to discuss evolutionary computation methods for generating algorithms and/or hyper-heuristics. These methods have the advantage of producing solutions that are applicable to any instance of a problem domain, instead of a solution specifically produced to a single instance of the problem. The areas of application of these methods may include, for instance, data mining, machine learning, and optimization.
[1] G. L. Pappa and A. A. Freitas, Automating the Design of Data Mining Algorithms: An Evolutionary Computation Approach, Springer, Natural Computing Series, 2010. xiii + 187 pages.
[2] E. k. Burke, M. Hyde, G. Kendall and J. Woodward, A genetic programming hyper-heuristic approach for evolving two dimensional strip packing heuristics. In: IEEE Transactions on Evolutionary Computation, 2010.
[3] 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.