The scalability problem is a major impediment to the use of hardware evolution for real-world circuit design problems. A potential solution is to model the map between genotype and phenotype on biological development. Although development has been shown to improve scalability for a few toy problems, it has not been demonstrated for any circuit design problems. This paper presents such a demonstration for two problems, the n-bit adder with carry and even n-bit parity problems, and shows that development imposes, and benefits from, fewer constraints on evolutionary innovation than other approaches to scalability. 1.

This paper discusses the underlying pressures responsible for code growth in genetic programming, and shows how an understanding of these pressures can be used to use to eliminate code growth while simultaneously improving performance. We begin with a discussion of two distinct components of code growth and the extent to which each component is relevant in practice. We then define the concept of resilience in GP trees, and show that the buildup of resilience is essential for code growth. We present simple modifications to the selection procedures used by GP that eliminate bloat without hurting performance. Finally, we show that eliminating bloat can improve the performance of genetic programming by a factor that increases as the problem is scaled in difficulty.

Abstract — Search algorithms are subject to the trappings of local optima. Attempts to address the problem are often framed in the context of needing to balance, or trade-off, exploitation against exploration. Ideally, it is best to maximise both simultaneously, but this is usually seen as infeasible in the presence of multi-modal search spaces. This paper investigates the potential for exploration of both neutrality and mutation rate, and argues that the former is the more important. The most interesting result, however, is that the necessity for a trade-off between exploitation and exploration can be avoided within the context of our algorithm for evolving Binary Decision Diagrams. I.

When a considerable number of mutations have no effects on fitness values, the fitness landscape is said neutral. In order to study the interplay between neutrality, which exists in many real-world applications, and performances of metaheuristics, it is useful to design landscapes which make it possible to tune precisely neutral degree distribution. Even though many neutral landscape models have already been designed, none of them are general enough to create landscapes with specific neutral degree distributions. We propose three steps to design such landscapes: first using an algorithm we construct a landscape whose distribution roughly fits the target one, then we use a simulated annealing heuristic to bring closer the two distributions and finally we affect fitness values to each neutral network. Then using this new family of fitness landscapes we are able to highlight the interplay between deceptiveness and neutrality.

Abstract. Prior program representations for genetic programming that incorporated features of modern programming languages solved harder problems than earlier representations, but required more complex genetic operators. We develop the idea of using combinator expressions as a program representation for genetic programming. This representation makes it possible to evolve programs with a variety of programming language constructs using simple genetic operators. We investigate the effort required to evolve combinator-expression solutions to several problems: linear regression, even parity on N inputs, and implementation of the stack and queue data structures. Genetic programming with combinator expressions compares favorably to prior approaches, namely the works

Search spaces sampled by the process of Genetic Programming often consist of programs which can represent a function in many different ways. Thus, when the space is examined it is highly likely that different programs may be tested which represent the same function, which is an undesirable waste of resources. It is argued that, if a search space can be constructed where only unique representations of a function are permitted, then this will be more successful than employing multiple representations. When the search space consists of canonical representations it is called a canonical search space, and when Genetic Programming is applied to this search space, it is called Canonical Representation Genetic Programming. The challenge lies in constructing these search spaces. With some function sets this is a trivial task, and with some function sets this is impossible to achieve. With other function sets it is not clear how the goal can be achieved. In this paper, we specifically examine the search space defined by the function set {+, −, ∗, /} and the terminal set {x, 1}. Drawing inspiration from the fundamental theorem of arithmetic, and results regarding the fundamental theorem of algebra, we construct a representation where each function that can be constructed with this primitive set has a unique representation.

Neutrality of some boolean parity fitness landscapes is investigated in this paper. Compared with some well known contributions on the same issue, we define some new measures that help characterizing neutral landscapes, we use a new sampling methodology, which captures some features that are disregarded by uniform random sampling, and we introduce new genetic operators to define the neighborhood of tree structures. We compare the fitness landscape induced by two different sets of functional operators ({nand} and {xor; not}). The different characteristics of the neutral networks seem to justify the different difficulties of these landscapes for genetic programming.

A coevolutionary competitive learning environment for two antagonistic agents is presented. The agents are controlled by a new kind of computational network based on a compartmentalised model of neurons. The genetic basis of neurons is an important [27] and neglected aspect of previous approaches. Accordingly, we have defined a collection of chromosomes representing various aspects of the neuron: soma, dendrites and axon branches, and synaptic connections. Chromosomes are represented and evolved using a form of genetic programming (GP) known as Cartesian GP. The network formed by running the chromosomal programs, has a highly dynamic morphology in which neurons grow, and die, and neurite branches together with synaptic connections form and change in response to environmental interactions. The idea of this paper is to demonstrate the importance of the genetic transfer of learned experience and life time learning. The learning is a consequence of the complex dynamics produced as a result of interaction (coevolution) between two intelligent agents. Our results show that both agents exhibit interesting learning capabilities. Categories and Subject Descriptors

The brain has long been seen as a powerful analogy from which novel computational techniques could be devised. However, most artificial neural network approaches have ignored the genetic basis of neural functions. In this paper we describe a radically different approach. We have devised a compartmental model of a neuron as a collection of seven chromosomes encoding distinct computational functions representing aspects of real neurons. This model allows neurons, dendrites, and axon branches to grow, die and change while solving a computational problem. This also causes the synaptic morphology to change and affect the information processing. Since the appropriate computational equivalent functions of neural computation are unknown, we have used a form of genetic programming known as Cartesian Genetic Programming (CGP) to obtain these functions. We have evaluated the learning potential of this system in the context of solving a well known agent based learning scenario, known as wumpus world and obtained promising results. Categories and Subject Descriptors

Abstract. Over the last years, neutrality has inspired many researchers in the area of Evolutionary Computation (EC) systems in the hope that it can aid evolution. However, there are contradictory results on the effects of neutrality in evolutionary search. The aim of this paper is to understand how neutrality-named in this paper degree neutrality- affects GP search. For analysis purposes, we use a well-defined measure of hardness (i.e., fitness distance correlation) as an indicator of difficulty in the absence and in the presence of neutrality, we propose a novel approach to normalise distances between a pair of trees and finally, we use a problem with deceptive features where GP is well-known to have poor performance and see the effects of neutrality in GP search. 1