Causality relates changes in the structure of an object with the effects of such changes, that is changes in the properties or behavior of the object. This paper analyzes the concept of causality in Genetic Programming (GP) and suggests how it can be used in adapting control parameters for speeding up GP search. We first analyze the effects of crossover to show the weak causality of the GP representation and operators. Hierarchical GP approaches based on the discovery and evolution of functions amplify this phenomenon. However, selection gradually retains strongly causal changes. Causality is correlated to search space exploitation and is discussed in the context of the exploration-exploitation tradeoff. The results described argue for a bottom-up GP evolutionary thesis. Finally, new developments based on the idea of GP architecture evolution (Koza, 1994a) are discussed from the causality perspective. Proceedings of the Fifth International Conference (ICGA95) Morgan Kaufmann, San Franc...

Hierarchical genetic programming (HGP) approaches rely on the discovery, modification, and use of new functions to accelerate evolution. This paper provides a qualitative explanation of the improved behavior of HGP, based on an analysis of the evolution process from the dual perspective of diversity and causality. From a static point of view, the use of an HGP approach enables the manipulation of a population of higher diversity programs. Higher diversity increases the exploratory ability of the genetic search process, as demonstrated by theoretical and experimental fitness distributions and expanded structural complexity of individuals. From a dynamic point of view, an analysis of the causality of the crossover operator suggests that HGP discovers and exploits useful structures in a bottom-up, hierarchical manner. Diversity and causality are complementary, affecting exploration and exploitation in genetic search. Unlike other machine learning techniques that need extra machinery to co...

Hierarchical genetic programming (HGP) approaches rely on the discovery, modification, and use of new functions to accelerate evolution. This paper provides a qualitative explanation of the improved behavior of HGP, based on an analysis of the evolution process from the dual perspective of diversity and causality. From a static point of view, the use of an HGP approach enables the manipulation of a population of higher diversity programs. Higher diversity increases the exploratory ability of the genetic search process, as demonstrated by theoretical and experimental fitness distributions and expanded structural complexity of individuals. From a dynamic point of view, this report analyzes the causality of the crossover operator. Causality relates changes in the structure of an object with the effect of such changes, i.e. changes in the properties or behavior of the object. The analyses of crossover causality suggests that HGP discovers and exploits useful structures in a bottom-up, hier...

The appropriate choice of the genotype ! phenotype mapping in combination with the mutation operator is important for a successful evolutionary search process. We suggest a measure to quantify the quality of this combination by addressing the question whether the relation among distances is carried over from one space to the other. Search processes which do not destroy the neighbourhood structure are termed strongly causal. We apply the proposed measure to parameter and structure optimisation problems in order to assess the combination (mapping, mutation operator) and at the same time to be able to propose improved settings. 1 Introduction The optimisation process in evolutionary algorithms is largely influenced by the mapping from the genotype space to the phenotype space. Especially for structure optimisation problems a measure of the quality of the combination (mapping, mutation, crossover) would be desirable. In this paper we propose such a measure based upon the ob...

I would like to thank my two advisors, Dr. Robert Sturges, Jr. and Dr. William “Red ” Whittaker for guiding my research. Bob has provided key technical insights, has critically read through severals drafts of this work and has made himself readily accessible. Red has provided both the finincial backing for this work and a far-reaching vision of the future of robotics and this work. My thesis owes much to this unique combination of advisors. I would also like to thank the other members of my dissertation committee, Dr. Dwight Baumann, Dr. Jonathan Cagan and Mr. John Wiss for their technical counsel and critical reviews of this work. I would also like to thank Dr. Subhas Desa for starting me down the road that has led to this thesis. I would not have been able to have done this work without the support of the CMU planetary rover group. This group has both generously given of their time and assistance. In addition, it was through the needs of this group that the need for this work first became apparent. In particular, I would like to thank Dr. Eric Krotkov, Dr. Reid Simmons and Mr. Kevin Dowling for their support. Much of this work was done under NASA contracts to CMU. I wish to thank Mr.

The main focus of the optimization of artificial neural networks has been the design of a problem dependent network structure in order to reduce the model complexity and to minimize the model error. Driven by a concrete application we identify in this paper another desirable property of neural networks -- the ability of the network to efficiently solve related problems denoted as a class of problems. In a more theoretical framework the aim is to develop neural networks for adaptability -- networks that learn (during evolution) to learn (during operation) . Evolutionary algorithms have turned out to be a robust method for the optimization of neural networks. As this process is time consuming, it is therefore also from the perspective of efficiency desirable to design structures that are applicable to many related problems. In this paper, two different approaches to solve this problem are studied, called ensemble method and generation method. We empirically show that an averaged Lamarcki...

We present an application of artificial neural networks to machine condition monitoring. Since several signal preprocessing methods produce high dimensional feature vectors there is a need for optimizing the structure of the neural network. We examine three methods for finding the structure and compare the resulting networks: fully connected networks, pruning techniques and "Structure Evolution and Incomplete Induction". Keywords: sound analysis, estimation, structure evolution 1 Introduction Machine condition monitoring is an area of growing interest for manufacturing and operation. In order to reduce maintenance costs it is necessary to detect mechanical failures using a monitoring system. Since malfunctions often produce extraordinary vibrations which could be detected by a human supervisor acoustic monitoring is a widely used technique. Especially for methods like nonparametric spectral estimation and time-frequency-analysis there is a need for pattern recognition methods wh...

The advent of Grid Computing promises users the power to have access to a vast amount of heterogeneous, distributed resources. The envisaged goal is to enable users and applications to seamlessly access these resources to solve complex large-scale problems whether in science, engineering, or commerce. To realize this goal the numerous barriers that normally separate different computing systems within and across organizations must be addressed. This can be achieved through effective standardization. Standardization plays a crucial role in achieving interoperability, portability and reusability of components and systems. The Open Grid Services Infrastructure (OGSI) Specification defines a set of conventions and extensions that contribute to the standardization of Grid Services. Since Grid Services are typically implemented as Web Services, they rely heavily on XML technologies. Thus, the focus of this research is to propose a multilaterally-interoperable Grid Service Framework that is robust and secure, using common XML standards relating to SOAP and Web Services. (After only four months into the research, only an analysis of the problem context is provided).

Abstract not found

The appropriate choice of the genotype → phenotype mapping in combination with the mutation operator is important for a successful evolutionary search process. We suggest a measure to quantify the quality of this combination by addressing the question whether the relation among distances is carried over from one space to the other. Search processes which do not destroy the neighbourhood structure are termed strongly causal. We apply the proposed measure to parameter and structure optimisation problems in order to assess the combination (mapping, mutation operator) and at the same time to be able to propose improved settings. 1