Evelyne Lutton

Parallel Problem Solving from Nature PPSN VI

Spatially structured evolutionary algorithms (EAs) have shown to be endowed with useful features for global optimization. Distributed EAs (dEA) and cellular EAs (cEA) are two of the most widely known types of structured algorithms. In this paper we deal with cellular EAs. Two important parameters guiding the search in a cEA are the population topology and the neighborhood defined on it. Here we first review some theoretical results which show that a cEA with a 2D grid can be easily tuned to shift from exploration to exploitation. We initially make a study on the relationship between the topology and the neighborhood by defining a ratio measure between they two. Then, we encompass a set of tests aimed at discovering the performance that different ratio values have on different classes of problems. We find out that, with the same neighborhood, rectangular grids have some advantages in multimodal and epistatic problems, while square ones are more efficient for solving deceptive problems and for simple function optimization. Finally, we propose and study a cEA in which the ratio is dynamically changed.

Take It EASEA

Evolutionary algorithms are not straightforward to implement and the lack of any specialised language forces users to reinvent the wheel every time they want to write a new program. Over the last years, evolutionary libraries have appeared, trying to reduce the amount of work involved in writing such algorithms from scratch, by offering standard engines, strategies and tools. Unfortunately, most of these libraries are quite complex to use, and imply a deep knowledge of object programming and C++. To further reduce the amount of work needed to implement a new algorithm, without however throwing down the drain all the man-years already spent in the development of such libraries, we have designed EASEA (acronym for EAsy Specification of Evolutionciry Algorithms): a new high-level language dedicated to the specification of evolutionary algorithms. EASEA compiles .ez files into source files in a target language, containing function calls to a chosen existing library. The resulting source file is in turn compiled and linked with the library to produce an executable file implementing the evolutionary algorithm specified in the original .ez file. EASEA vO.4 is available at: http://www-rocq.inria.fr/EVO-Lab/.

Compact Unstructured Representations for Evolutionary Design

This paper proposes a few steps to escape structured extensive representations for objects, in the context of evolutionary Topological Optimum Design (TOD) problems: early results have demonstrated the potential power of Evolutionary methods to find numerical solutions to yet unsolved TOD problems, but those approaches were limited because the complexity of the representation was that of a fixed underlying mesh. Different compact unstructured representations are introduced, the complexity of which is self-adaptive, i.e. is evolved by the algorithm itself. The Voronoi-based representations are variable length lists of alleles that are directly decoded into object shapes, while the IFS representation, based on fractal theory, involves a much more complex morphogenetic process. First results demonstrates that Voronoi-based representations allow one to push further the limits of Evolutionary Topological Optimum Design by actually removing the correlation between the complexity of the representations and that of the discretization. Further comparative results among all these representations on simple test problems seem to indicate that the complex causality in the IFS representation disfavors it compared to the Voronoi-based representations.

Polar IFS+Parisian Genetic ProgrammingeEfficient IFS Inverse Problem Solving

This paper proposes a new method for treating the inverse problem for Iterated Functions Systems (IFS) using Genetic Programming. This method is based on two original aspects. On the fractal side, a new representation of the IFS functions, termed Polar Iterated Functions Systems, is designed, shrinking the search space to mostly contractive functions. Moreover, the Polar representation gives direct access to the fixed points of the functions. On the evolutionary side, a new variant of GP, the “Parisian” approach is presented. The paper explains its similarity to the “Michigan” approach of Classifier Systems: each individual of the population only represents a part of the global solution. The solution to the inverse problem for IFS is then built from a set of individuals. A local contribution to the global fitness of an IFS is carefully defined for each one of its member functions and plays a major role in the fitness of each individual. It is argued here that both proposals result in a large improvement in the algorithms. We observe a drastic cut-down on CPU-time, obtaining good results with small populations in few generations.

A genetic algorithm for the detection of 2D geometric primitives in images

We investigate the use of genetic algorithms (GAs) for image primitives extraction (such as segments, circles, ellipses or quadrilaterals). This approach completes the well-known Hough transform, in the sense that GAs are efficient when the Hough approach becomes too expensive in memory, i.e. when we search for complex primitives having more than 3 or 4 parameters. A GA is a stochastic technique, relatively slow, but which provides with an efficient tool to search in a high dimensional space. The philosophy of the method is very similar to the Hough transform, which is to search an optimum in a parameter space. However, we will see that the implementation is different.

FRACTAL MODELING OF SPEECH SIGNALS

In this paper, we present a method for speech signal analysis and synthesis based on IFS theory. We consider a speech signal as the graph of a continuous function whose irregularity, measured in terms of its local Holder exponents, is arbitrary. We extract a few remarkable points in the signal and perform a fractal interpolation between them using a classical technique based on IFS theory. We thus obtain a functional representation of the speech signal, which is well adapted to various applications, as for instance voice interpolation.

Parisian camera placement for vision metrology

This paper presents a novel camera network design methodology based on the Parisian evolutionary computation approach. This methodology proposes to partition the original problem into a set of homogeneous elements, whose individual contribution to the problem solution can be evaluated separately. A population comprised of these homogeneous elements is evolved with the goal of creating a single solution by a process of aggregation. The goal of the Parisian evolutionary process is to locally build better individuals that jointly form better global solutions. The implementation of the proposed approach requires addressing aspects such as problem decomposition and representation, local and global fitness integration, as well as diversity preservation mechanisms. The benefit of applying the Parisian approach to our camera placement problem is a substantial reduction in computational effort expended in the evolutionary optimization process. Moreover, experimental results coincide with previous state of the art photogrammetric network design methodologies, while incurring in only a fraction of the computational cost.

Preface: Introduction to the special issue on evolutionary computer vision and image understanding

Genetic and Evolutionary Computation (GEC) is a recent research field in computer science which deals with adaptive systems and optimization techniques inspired by the rules of natural evolution. One of its goals is to endow computers with information-processing capabilities comparable to those found in nature (Holland, 1992; Poli and Cagnoni, 2003; Koza, 1992; Schwefel, 1981; Mitchell, 1996; Landon and Poli, 2002; Goldberg, 1989). The general applicability of its methods makes it possible to use GEC to solve problems in a large number of applications. In particular, GEC methods can be applied effectively to those fields whose tasks require robust and flexible techniques to optimize performance in the many possible scenarios that characterize real-world problems (GECCO; CEC; PPSN; EuroGP). Among those fields, computer vision and image understanding (CVIU) represents one of the most challenging for the complexity of the tasks that are being solved in order to provide computers with human-like perception capabilities, allowing them to sense the environment, understand the sensed data, identify patterns, take appropriate actions and learn from experience to enhance future performance (CVPR; ICCV; ECCV; ICPR). Real-world applications of CVIU presently include autonomous robot or vehicle navigation, inspection, quality control, surveillance, to mention but a few. To achieve these high-level tasks, lower-level problems need to be solved, such as feature extraction, 3D modeling, and object classification. These real-world tasks require to be robust and flexible to optimize performance in diverse scenarios encountered in a given application. CVIU is steadily gaining relevance within the large number of application fields of GEC techniques, thanks to the capability of the latter to explore huge search domains effectively, searching and often finding solutions that lie well far away from the rather limited region spanned by more traditional, hand-coded ones. A first benefit of studying GEC techniques within the computational CVIU framework is to mature the information-processing capabilities of artificial systems based on challenging real-world problems. A second benefit is the promise of advancing the CVIU techniques with a bet-

Speciation in Behavioral Space for Evolutionary Robotics

In Evolutionary Robotics, population-based evolutionary computation is used to design robot neurocontrollers that produce behaviors which allow the robot to fulfill a user-defined task. However, the standard approach is to use canonical evolutionary algorithms, where the search tends to make the evolving population converge towards a single behavioral solution, even if the high-level task could be accomplished by structurally different behaviors. In this work, we present an approach that preserves behavioral diversity within the population in order to produce a diverse set of structurally different behaviors that the robot can use. In order to achieve this, we employ the concept of speciation, where the population is dynamically subdivided into sub-groups, or species, each one characterized by a particular behavioral structure that all individuals within that species share. Speciation is achieved by describing each neurocontroller using a representations that we call a behavior signature, these are descriptors that characterize the traversed path of the robot within the environment. Behavior signatures are coded using character strings, this allows us to compare them using a string similarity measure, and three measures are tested. The proposed behavior-based speciation is compared with canonical evolution and a method that speciates based on network topology. Experimental tests were carried out using two robot tasks (navigation and homing behavior), several training environments, and two different robots (Khepera and Pioneer), both real and simulated. Results indicate that behavior-based speciation increases the diversity of the behaviors based on their structure, without sacrificing performance. Moreover, the evolved controllers exhibit good robustness when the robot is placed within environments that were not used during training. In conclusion, the speciation method presented in this work allows an evolutionary algorithm to produce several robot behaviors that are structurally different but all are able to solve the same robot task.

Discovering several robot behaviors through speciation

This contribution studies speciation from the standpoint of evolutionary robotics (ER). A common approach to ER is to design a robots control system using neuro-evolution during training. An extension to this methodology is presented here, where speciation is incorporated to the evolution process in order to obtain a varied set of solutions for a robotics problem using a single algorithmic run. Although speciation is common in evolutionary computation, it has been less explored in behavior-based robotics. When employed, speciation usually relies on a distance measure that allows different individuals to be compared. The distance measure is normally computed in objective or phenotypic space. However, the speciation process presented here is intended to produce several distinct robot behaviors; hence, speciation is sought in behavioral space. Thence, individual neurocontrollers are described using behavior signatures, which represent the traversed path of the robot within the training environment and are encoded using a character string. With this representation, behavior signatures are compared using the normalized Levenshtein distance metric (N-GLD). Results indicate that speciation in behavioral space does indeed allow the ER system to obtain several navigation strategies for a common experimental setup. This is illustrated by comparing the best individual from each species with those obtained using the Neuro-Evolution of Augmenting Topologies (NEAT) method which speciates neural networks in topological space.