Automated Design of Probability Distributions as Mutation Operators for Evolutionary Programming Using Genetic Programming

In this study we use Genetic Programming (GP) as an offline hyper-heuristic to evolve a mutation operator for Evolutionary Programming. This is done using the Gaussian and uniform distributions as the terminal set, and arithmetic operators as the function set. The mutation operators are automatically designed for a specific function class. The contribution of this paper is to show that a GP can not only automatically design a mutation operator for Evolutionary Programming (EP) on functions generated from a specific function class, but also can design more general mutation operators on functions generated from groups of function classes. In addition, the automatically designed mutation operators also show good performance on new functions generated from a specific function class or a group of function classes.

In this paper, we discuss the implementation of Evolutionary Programming (EP) using a mutation operator based on Levy probability distribution and investigate its performance in optimization of selected antenna problems. Since the shape-parameter, alpha, in the Levy distribution results in Cauchy and Gaussian distributions for values of 1 and 2, respectively, one can design adaptive schemes that dynamically change the shape of Levy distribution in generation of offsprings during the evolutionary process in order to exploit the desirable properties of these two distributions. In particular, we introduce a simple adaptive scheme and compare its performance in antenna optimization to adaptive and non-adaptive schemes, previously reported in the literature.

Conventional implementation of evolutionary programming (EP) for continuous parameter optimization uses Gaussian mutations. However, an implementation of EP with Cauchy mutation operator has empirically been shown to outperform EP using the Gaussian mutations for optimizations of multi-modal functions with many local optima. The faster convergence of EP with Cauchy mutation has been explained in terms of the fatter tails of the Cauchy probability density function, which results in a higher probability of escaping from a local optimum. In this work we present an implementation of EP consisting of a hybrid linear combination of the Cauchy and Gaussian mutations for antenna optimization problems in order to exploit the desirable properties of these two operators. The implementation follows a procedure similar to the one described by Chellapilla (see IEEE Trans. on Evolutionary Computation, September 1998) for function optimization and uses a self-adaptive scheme for updating the standard deviation and the scale parameter of Gaussian and Cauchy distributions, respectively, during the evolution. As an example the optimization of a six-element Yagi-Uda array of dipoles using the hybrid method is presented.

In this paper we address the role of divergence and convergence in creative processes, and argue about the need to consider them in Computational Creativity research in the Genetic or Evolutionary Programming paradigm, being one's goal the problem of the Halt Condition in Genetic Programming. Here the candidate solutions are seen as evolutionary logic programs or theories, being the test whether a solution is optimal based on a measure of the quality-of-information carried out by those logical theories or programs. Furthermore, we present Conceptual Blending Theory as being a promising framework for implementing convergence methods within creativity programs, in terms of the logic programming framework.

March 1-3, 1995, San Diego, California Evolutionary programming is one of the predominate algorithms withing the rapidly expanding field of evolutionary computation. These edited contributions to the Fourth Annual Conference on Evolutionary Programming are by leading scientists from academia, industry, and defense. The papers describe both the theory and practical application of evolutionary programming, as well as other methods of evolutionary computation including evolution strategies, genetic algorithms, genetic programming, and cultural algorithms.Topics include :- Novel Areas of Evolutionary Programming and Evolution Strategies.- Evolutionary Computation with Medical Applications.- Issues in Evolutionary Optimization Pattern Discovery, Pattern Recognition, and System Identification.- Hierarchical Levels of Learning.- Self-Adaptation in Evolutionary Computation.- Morphogenic Evolutionary Computation.- Issues in Evolutionary Optimization.- Evolutionary Applications to VLSI and Part Placement.- Applications of Evolutionary Computation to Biology and Biochemistry Control.- Applications of Evolutionary Computation.- Genetic and Inductive Logic Programming.- Genetic Neural Networks.- The Future of Evolutionary Computation.A Bradford Book. Complex Adaptive Systems series

March 1-3, 1995, San Diego, California Evolutionary programming is one of the predominate algorithms withing the rapidly expanding field of evolutionary computation. These edited contributions to the Fourth Annual Conference on Evolutionary Programming are by leading scientists from academia, industry, and defense. The papers describe both the theory and practical application of evolutionary programming, as well as other methods of evolutionary computation including evolution strategies, genetic algorithms, genetic programming, and cultural algorithms.Topics include :- Novel Areas of Evolutionary Programming and Evolution Strategies.- Evolutionary Computation with Medical Applications.- Issues in Evolutionary Optimization Pattern Discovery, Pattern Recognition, and System Identification.- Hierarchical Levels of Learning.- Self-Adaptation in Evolutionary Computation.- Morphogenic Evolutionary Computation.- Issues in Evolutionary Optimization.- Evolutionary Applications to VLSI and Part Placement.- Applications of Evolutionary Computation to Biology and Biochemistry Control.- Applications of Evolutionary Computation.- Genetic and Inductive Logic Programming.- Genetic Neural Networks.- The Future of Evolutionary Computation.A Bradford Book. Complex Adaptive Systems series

Evolutionary Programming (EP) and Differential Evolution (DE) are well known as simple and efficient schemes for global optimization over continuous spaces. Both EP and DE use mutation for producing offspring. The mutation operators of EP usually generate the search step size for mutation by probability distribution functions while the mutation operators of DE generate it by adding a weighted difference vector between two individuals to a third individual. In this paper, a new EP algorithm is proposed based on dual mutation strategy (DMEP) as it incorporates both the mutation operators of EP and DE literature. Thus the balance between exploration and exploitation is obtained by two different categories of mutation operators. To evaluate the performance of the proposed scheme, a test-suite of 37 benchmark functions has been used and results have been compared with some prominent evolutionary systems. Experimental results show the remarkable effectiveness of the dual mutation strategy employed by DMEP.

Genetic programming (GP) is an evolutionary algorithm that evolves computer programs. Its main recombination operator is standard one point crossover which is generally accepted to be one of GP's weak points, due to its ignorance of the context into which genetic material is placed. This work introduces a new context aware recombination operator called context-aware crossover. It implicitly calculates the best possible context of the subtree-to- be-exchanged in the other parent and places it there. It is tested on a wide range of problems and found quite constructive in general and quite effective on hard problems, in particular. It has also shown the ability to generate quite smaller trees than standard GP without effecting the fitness of a population adversely.

Reactive power optimisation is an important technique, which is concerned with the security and economy of operation of the power system. The appropriate distribution of reactive power can elevate voltage rating, decrease network losses and maintain network running under proper conditions. An improved evolutionary programming method with dynamic mutation and metropolis selection to solve the multi-objective reactive power optimisation under the deregulation environment is presented. The multi-objective function includes the minimisation of network losses, voltage deviation and compensation cost. To solve the problems of convergence and robustness in the conventional evolutionary programming method, the mutation operators and the selection criteria that affect the convergence and robustness are considered and a dynamic mutation and metropolis selection evolutionary programming method is suggested. Introducing chaos dynamics into mutation operators of evolutionary programming, the new method adopts the certainty method of like-stochastic to obtain mutation operators which break through the conventional thought with mutation by stochastic numbers of fixed distribution. Also, it introduces the metropolis selection in evolutionary programming to construct new selection operators. Thus, the new method not only accelerates convergence but also increases precision, so is an efficient way to optimise the capacitor banks and the adjustable transformer ratio. Tested by the IEEE-30 bus system, the method is effective.

Memory plays an important part in the development of the collective and individual fitness of a biological species. Knowledge acquired by teaching and experience is stored in memory and retrieved when an individual is required to perform a task or solve a problem. In this article we discuss the concept of memory in a genetic programming (GP) system. The memory operator we introduce is used in conjunction with, and in support of, the traditional selection, crossover, and mutation operators. We implement a form of memory by preserving the component parts of good designs from previous generations. The memory is used as a pool of genetic material which is introduced into later generations by a form of mutation. It thus acts as a ’tribal folklore’ or cultural memory for the system. Using the memory operator, significant improvements in performance can be achieved on some problems. We believe that these improvements may be achievable in most areas in which Genetic Programming is applied, given suitable encodings, fitness measures and parameter tuning.

Semantic genetic programming is a recent, rapidly growing trend in Genetic Programming (GP) that aims at opening the 'black box' of the evaluation function and make explicit use of more information on program behavior in the search. In the most common scenario of evaluating a GP program on a set of input-output examples (fitness cases), the semantic approach characterizes program with a vector of outputs rather than a single scalar value (fitness). The past research on semantic GP has demonstrated that the additional information obtained in this way facilitates designing more effective search operators. In particular, exploiting the geometric properties of the resulting semantic space leads to search operators with attractive properties, which have provably better theoretical characteristics than conventional GP operators. This in turn leads to dramatic improvements in experimental comparisons. The aim of the tutorial is to give a comprehensive overview of semantic methods in genetic programming, illustrate in an accessible way a formal geometric framework for program semantics to design provably good mutation and crossover operators for traditional GP problem domains, and to analyze rigorously their performance (runtime analysis). A number of real-world applications of this framework will be also presented. Other promising emerging approaches to semantics in GP will be reviewed. In particular, the recent developments in the behavioral programming, which aims at characterizing the entire program behavior (and not only program outputs) will be covered as well. Current challenges and future trends in semantic GP will be identified and discussed. Selected methods and concepts will be accompanied with live software demonstrations. Also, efficient implementation of semantic search operators may be challenging. We will illustrate very efficient, concise and elegant implementations of these operators, which are available for download from the web.

This paper presents an approach for designing classifiers for a multiclass problem using Genetic Programming (GP). The proposed approach takes an integrated view of all classes when GP evolves. An individual of the population will be represented using multiple trees. The GP is trained with a set of N training samples in steps. A concept of unfitness of a tree is used in order to improve genetic evolution. Weak trees having poor performance are given more chance to participate in the genetic operations, and thus improve themselves. In this context, a new mutation operation called nondestructive directed point mutation is used, which reduces the destructive nature of mutation operation. The approach is being demonstrated by experimenting on some datasets.

In this paper, we provide theoretical bases for the use of two classes of evolutionary computation, that is evolutionary programming and genetic programming, that support automated inference of fuzzy automaton-driven parser of natural language. This parser, called fPDAMS (fuzzy nondeterministic pushdown automaton with associative memory access), works in the stratificational knowledge representation system (SKRS system), which is an attempt at formalizing the multi-layer structure of natural language. In the evolutionary programming (EP) each chromosome is a representation of a transition graph of fPDAMS. The key mechanism used in this approach is asexual mutation. Genetic programing (GP) can be applied to natural language parser if a mapping is established between the point-labeled tree used in GP and the transition graph of the automaton. The method of mapping is based on the encoding technique called “Cellular Encoding” and proposed by F.Gruau.

By synthesizing a growing body ofwork in search processes that are not driven by explicit objectives, this paper advances the hypothesis that there is a fundamental problem with the dominant paradigm of objective-based search in evolutionary computation and genetic programming: Most ambitious objectives do not illuminate a path to themselves. That is, the gradient of improvement induced by ambitious objectives tends to lead not to the objective itself but instead to deadend local optima. Indirectly supporting this hypothesis, great discoveries often are not the result of objective-driven search. For example, the major inspiration for both evolutionary computation and genetic programming, natural evolution, innovates through an open-ended process that lacks a final objective. Similarly, large-scale cultural evolutionary processes, such as the evolution of technology, mathematics, and art, lack a unified fixed goal. In addition, direct evidence for this hypothesis is presented from a recently-introduced search algorithm called novelty search. Though ignorant of the ultimate objective of search, in many instances novelty search has counter-intuitively outperformed searching directly for the objective, including a wide variety of randomly-generated problems introduced in an experiment in this chapter. Thus a new understanding is beginning to emerge that suggests that searching for a fixed objective, which is the reigning paradigm in evolutionary computation and even machine learning as a whole, may ultimately limit what can be achieved. Yet the liberating implication of this hypothesis argued in this paper is that by embracing search processes that are not driven by explicit objectives, the breadth and depth of what is reachable through evolutionary methods such as genetic programming may be greatly expanded.

The Identification of rock constitutive model is a typical nonlinear system problem; introduction of the intelligent methods has greatly stimulated the research. In this paper, genetic algorithm , BP neural network and genetic programming are used in the identification of constitutive model ,the capacity of three intelligent methods in the model identification is compared ,whose results show that genetic algorithm can largely reduce the probability of trapping into localized optimum solution by means of its global searching ability ,however , because its pre-assumption of constitutive model has already lead to some error ,its fitting results are not so good . BP neural network and genetic programming need not pre-assume the structure of constitutive model ,the nonlinear mapping expressions between stress and strain are obtained by self-organization and self-learning .But ,the determination of BP neural network structure is experience-depending and time-consuming . In addition, genetic programming need not define its structure, so long as the functional set and terminal set are chosen, after series of genetic operations, the nonlinear relations between variables are outputted, therefore, the operation of genetic programming is simple which realize the visualization of the constitutive equations.