arXiv subjects

Marc Schoenauer

Publications and source records attributed to Marc Schoenauer.

At least 73 records · Page 4Linked to original sources

arXiv2012

Self-Adaptive Surrogate-Assisted Covariance Matrix Adaptation Evolution Strategy

This paper presents a novel mechanism to adapt surrogate-assisted population-based algorithms. This mechanism is applied to ACM-ES, a recently proposed surrogate-assisted variant of CMA-ES. The resulting algorithm, saACM-ES, adjusts online the lifelength of the current surrogate model (the number of CMA-ES generations before learning a new surrogate) and the surrogate hyper-parameters. Both heuristics significantly improve the quality of the surrogate model, yielding a significant speed-up of saACM-ES compared to the ACM-ES and CMA-ES baselines. The empirical validation of saACM-ES on the BBOB-2012 noiseless testbed demonstrates the efficiency and the scalability w.r.t the problem dimension and the population size of the proposed approach, that reaches new best results on some of the benchmark problems.

cs.NE↗

arXiv2010

Open-Ended Evolutionary Robotics: an Information Theoretic Approach

This paper is concerned with designing self-driven fitness functions for Embedded Evolutionary Robotics. The proposed approach considers the entropy of the sensori-motor stream generated by the robot controller. This entropy is computed using unsupervised learning; its maximization, achieved by an on-board evolutionary algorithm, implements a "curiosity instinct", favouring controllers visiting many diverse sensori-motor states (sms). Further, the set of sms discovered by an individual can be transmitted to its offspring, making a cultural evolution mode possible. Cumulative entropy (computed from ancestors and current individual visits to the sms) defines another self-driven fitness; its optimization implements a "discovery instinct", as it favours controllers visiting new or rare sensori-motor states. Empirical results on the benchmark problems proposed by Lehman and Stanley (2008) comparatively demonstrate the merits of the approach.

cs.RO↗

arXiv2010

Experimental Comparisons of Derivative Free Optimization Algorithms

In this paper, the performances of the quasi-Newton BFGS algorithm, the NEWUOA derivative free optimizer, the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), the Differential Evolution (DE) algorithm and Particle Swarm Optimizers (PSO) are compared experimentally on benchmark functions reflecting important challenges encountered in real-world optimization problems. Dependence of the performances in the conditioning of the problem and rotational invariance of the algorithms are in particular investigated.

math.NA↗

arXiv2010

Evolving Genes to Balance a Pole

We discuss how to use a Genetic Regulatory Network as an evolutionary representation to solve a typical GP reinforcement problem, the pole balancing. The network is a modified version of an Artificial Regulatory Network proposed a few years ago, and the task could be solved only by finding a proper way of connecting inputs and outputs to the network. We show that the representation is able to generalize well over the problem domain, and discuss the performance of different models of this kind.

cs.AI↗

arXiv2009

French Roadmap for complex Systems 2008-2009

This second issue of the French Complex Systems Roadmap is the outcome of the Entretiens de Cargese 2008, an interdisciplinary brainstorming session organized over one week in 2008, jointly by RNSC, ISC-PIF and IXXI. It capitalizes on the first roadmap and gathers contributions of more than 70 scientists from major French institutions. The aim of this roadmap is to foster the coordination of the complex systems community on focused topics and questions, as well as to present contributions and challenges in the complex systems sciences and complexity science to the public, political and industrial spheres.

nlin.AO↗

arXiv2007

GUIDE: Unifying Evolutionary Engines through a Graphical User Interface

Many kinds of Evolutionary Algorithms (EAs) have been described in the literature since the last 30 years. However, though most of them share a common structure, no existing software package allows the user to actually shift from one model to another by simply changing a few parameters, e.g. in a single window of a Graphical User Interface. This paper presents GUIDE, a Graphical User Interface for DREAM Experiments that, among other user-friendly features, unifies all kinds of EAs into a single panel, as far as evolution parameters are concerned. Such a window can be used either to ask for one of the well known ready-to-use algorithms, or to very easily explore new combinations that have not yet been studied. Another advantage of grouping all necessary elements to describe virtually all kinds of EAs is that it creates a fantastic pedagogic tool to teach EAs to students and newcomers to the field.

cs.NE↗

arXiv2007

Optimising the topology of complex neural networks

In this paper, we study instances of complex neural networks, i.e. neural netwo rks with complex topologies. We use Self-Organizing Map neural networks whose n eighbourhood relationships are defined by a complex network, to classify handwr itten digits. We show that topology has a small impact on performance and robus tness to neuron failures, at least at long learning times. Performance may howe ver be increased (by almost 10%) by artificial evolution of the network topo logy. In our experimental conditions, the evolved networks are more random than their parents, but display a more heterogeneous degree distribution.

cs.NE↗

arXiv2007

Identification of the Isotherm Function in Chromatography Using CMA-ES

This paper deals with the identification of the flux for a system of conservation laws in the specific example of analytic chromatography. The fundamental equations of chromatographic process are highly non linear. The state-of-the-art Evolution Strategy, CMA-ES (the Covariance Matrix Adaptation Evolution Strategy), is used to identify the parameters of the so-called isotherm function. The approach was validated on different configurations of simulated data using either one, two or three components mixtures. CMA-ES is then applied to real data cases and its results are compared to those of a gradient-based strategy.

math.OC↗

arXiv2007

Evolutionary Mesh Numbering: Preliminary Results

Mesh numbering is a critical issue in Finite Element Methods, as the computational cost of one analysis is highly dependent on the order of the nodes of the mesh. This paper presents some preliminary investigations on the problem of mesh numbering using Evolutionary Algorithms. Three conclusions can be drawn from these experiments. First, the results of the up-to-date method used in all FEM softwares (Gibb's method) can be consistently improved; second, none of the crossover operators tried so far (either general or problem specific) proved useful; third, though the general tendency in Evolutionary Computation seems to be the hybridization with other methods (deterministic or heuristic), none of the presented attempt did encounter any success yet. The good news, however, is that this algorithm allows an improvement over the standard heuristic method between 12% and 20% for both the 1545 and 5453-nodes meshes used as test-bed. Finally, some strange interaction between the selection scheme and the use of problem specific mutation operator was observed, which appeals for further investigation.

math.NA↗

arXiv2007

Evolving Symbolic Controllers

The idea of symbolic controllers tries to bridge the gap between the top-down manual design of the controller architecture, as advocated in Brooks' subsumption architecture, and the bottom-up designer-free approach that is now standard within the Evolutionary Robotics community. The designer provides a set of elementary behavior, and evolution is given the goal of assembling them to solve complex tasks. Two experiments are presented, demonstrating the efficiency and showing the recursiveness of this approach. In particular, the sensitivity with respect to the proposed elementary behaviors, and the robustness w.r.t. generalization of the resulting controllers are studied in detail.

cs.AI↗

arXiv2007

Robust Multi-Cellular Developmental Design

This paper introduces a continuous model for Multi-cellular Developmental Design. The cells are fixed on a 2D grid and exchange "chemicals" with their neighbors during the growth process. The quantity of chemicals that a cell produces, as well as the differentiation value of the cell in the phenotype, are controlled by a Neural Network (the genotype) that takes as inputs the chemicals produced by the neighboring cells at the previous time step. In the proposed model, the number of iterations of the growth process is not pre-determined, but emerges during evolution: only organisms for which the growth process stabilizes give a phenotype (the stable state), others are declared nonviable. The optimization of the controller is done using the NEAT algorithm, that optimizes both the topology and the weights of the Neural Networks. Though each cell only receives local information from its neighbors, the experimental results of the proposed approach on the 'flags' problems (the phenotype must match a given 2D pattern) are almost as good as those of a direct regression approach using the same model with global information. Moreover, the resulting multi-cellular organisms exhibit almost perfect self-healing characteristics.

cs.AI↗

arXiv2007

Ensemble Learning for Free with Evolutionary Algorithms ?

Evolutionary Learning proceeds by evolving a population of classifiers, from which it generally returns (with some notable exceptions) the single best-of-run classifier as final result. In the meanwhile, Ensemble Learning, one of the most efficient approaches in supervised Machine Learning for the last decade, proceeds by building a population of diverse classifiers. Ensemble Learning with Evolutionary Computation thus receives increasing attention. The Evolutionary Ensemble Learning (EEL) approach presented in this paper features two contributions. First, a new fitness function, inspired by co-evolution and enforcing the classifier diversity, is presented. Further, a new selection criterion based on the classification margin is proposed. This criterion is used to extract the classifier ensemble from the final population only (Off-line) or incrementally along evolution (On-line). Experiments on a set of benchmark problems show that Off-line outperforms single-hypothesis evolutionary learning and state-of-art Boosting and generates smaller classifier ensembles.

cs.AI↗

arXiv2006

On the Benefits of Inoculation, an Example in Train Scheduling

The local reconstruction of a railway schedule following a small perturbation of the traffic, seeking minimization of the total accumulated delay, is a very difficult and tightly constrained combinatorial problem. Notoriously enough, the railway company's public image degrades proportionally to the amount of daily delays, and the same goes for its profit! This paper describes an inoculation procedure which greatly enhances an evolutionary algorithm for train re-scheduling. The procedure consists in building the initial population around a pre-computed solution based on problem-related information available beforehand. The optimization is performed by adapting times of departure and arrival, as well as allocation of tracks, for each train at each station. This is achieved by a permutation-based evolutionary algorithm that relies on a semi-greedy heuristic scheduler to gradually reconstruct the schedule by inserting trains one after another. Experimental results are presented on various instances of a large real-world case involving around 500 trains and more than 1 million constraints. In terms of competition with commercial math ematical programming tool ILOG CPLEX, it appears that within a large class of instances, excluding trivial instances as well as too difficult ones, and with very few exceptions, a clever initialization turns an encouraging failure into a clear-cut success auguring of substantial financial savings.

cs.AI↗

arXiv2006

Genetic Programming for Kernel-based Learning with Co-evolving Subsets Selection

Support Vector Machines (SVMs) are well-established Machine Learning (ML) algorithms. They rely on the fact that i) linear learning can be formalized as a well-posed optimization problem; ii) non-linear learning can be brought into linear learning thanks to the kernel trick and the mapping of the initial search space onto a high dimensional feature space. The kernel is designed by the ML expert and it governs the efficiency of the SVM approach. In this paper, a new approach for the automatic design of kernels by Genetic Programming, called the Evolutionary Kernel Machine (EKM), is presented. EKM combines a well-founded fitness function inspired from the margin criterion, and a co-evolution framework ensuring the computational scalability of the approach. Empirical validation on standard ML benchmark demonstrates that EKM is competitive using state-of-the-art SVMs with tuned hyper-parameters.

cs.AI↗

arXiv2006

Using Domain Knowledge in Evolutionary System Identification

Two example of Evolutionary System Identification are presented to highlight the importance of incorporating Domain Knowledge: the discovery of an analytical indentation law in Structural Mechanics using constrained Genetic Programming, and the identification of the repartition of underground velocities in Seismic Prospection. Critical issues for sucessful ESI are discussed in the light of these results.

cs.AI↗

arXiv2006

Genetic Programming, Validation Sets, and Parsimony Pressure

Fitness functions based on test cases are very common in Genetic Programming (GP). This process can be assimilated to a learning task, with the inference of models from a limited number of samples. This paper is an investigation on two methods to improve generalization in GP-based learning: 1) the selection of the best-of-run individuals using a three data sets methodology, and 2) the application of parsimony pressure in order to reduce the complexity of the solutions. Results using GP in a binary classification setup show that while the accuracy on the test sets is preserved, with less variances compared to baseline results, the mean tree size obtained with the tested methods is significantly reduced.

cs.LG↗

arXiv2006

Divide-and-Evolve: a New Memetic Scheme for Domain-Independent Temporal Planning

An original approach, termed Divide-and-Evolve is proposed to hybridize Evolutionary Algorithms (EAs) with Operational Research (OR) methods in the domain of Temporal Planning Problems (TPPs). Whereas standard Memetic Algorithms use local search methods to improve the evolutionary solutions, and thus fail when the local method stops working on the complete problem, the Divide-and-Evolve approach splits the problem at hand into several, hopefully easier, sub-problems, and can thus solve globally problems that are intractable when directly fed into deterministic OR algorithms. But the most prominent advantage of the Divide-and-Evolve approach is that it immediately opens up an avenue for multi-objective optimization, even though the OR method that is used is single-objective. Proof of concept approach on the standard (single-objective) Zeno transportation benchmark is given, and a small original multi-objective benchmark is proposed in the same Zeno framework to assess the multi-objective capabilities of the proposed methodology, a breakthrough in Temporal Planning.

cs.AI↗

arXiv2005

Evolution of Voronoi based Fuzzy Recurrent Controllers

A fuzzy controller is usually designed by formulating the knowledge of a human expert into a set of linguistic variables and fuzzy rules. Among the most successful methods to automate the fuzzy controllers development process are evolutionary algorithms. In this work, we propose the Recurrent Fuzzy Voronoi (RFV) model, a representation for recurrent fuzzy systems. It is an extension of the FV model proposed by Kavka and Schoenauer that extends the application domain to include temporal problems. The FV model is a representation for fuzzy controllers based on Voronoi diagrams that can represent fuzzy systems with synergistic rules, fulfilling the $ε$-completeness property and providing a simple way to introduce a priory knowledge. In the proposed representation, the temporal relations are embedded by including internal units that provide feedback by connecting outputs to inputs. These internal units act as memory elements. In the RFV model, the semantic of the internal units can be specified together with the a priori rules. The geometric interpretation of the rules allows the use of geometric variational operators during the evolution. The representation and the algorithms are validated in two problems in the area of system identification and evolutionary robotics.

cs.AI↗