Genetic programming II, (1994)

by J R Koza
Results 11 - 20 of 81

A Controlled Experiment: Evolution for Learning Difficult Image Classification

by Astro Teller, Manuela Veloso - In Seventh Portuguese Conference On Artificial Intelligence
"... The signal-to-symbol problem is the task of converting raw sensor data into a set of symbols that Artificial Intelligence systems can reason about. Wc have developed a method for directly learning and combining algorithms that map signMs into symbols. This new method is based on evolutionary com ..."
Abstract - Cited by 20 (2 self) - Add to MetaCart
The signal-to-symbol problem is the task of converting raw sensor data into a set of symbols that Artificial Intelligence systems can reason about. Wc have developed a method for directly learning and combining algorithms that map signMs into symbols. This new method is based on evolutionary computation and imposes httle burden on or bias from the humans involved. Previous papers of ours have focused on PADO, our learning architecture. Wc showed how it apphcs to the general signal-to-symbol task and in particular the impressive results it brings to natural image object recognition. The most exciting challenge this work has received is the idea that PADO's success in natural image object recognition may be due to the underlying simphcity of the problems we posed it. This challenge imphcitly assumes that our approach suffers from many of the same mtictions that traditional computer vision approaches suffer in natural image object recognition. This paper responds to this challenge by designing and executing a controlled experiment specifically designed to sohdify PADO's clMm to success.
(Show Context)

SINERGY: A linear planner based on genetic programming

by Ion Muslea - Fourth European Conference on Planning, volume 1348 of Lecture notes in artificial intelligence , 1997
"... Abstract. In this paper we describe SINERGY, which is a highly parallelizable, linear planning system that is based on the genetic programming paradigm. Rather than reasoning about the world it is planning for, SINERGY uses artificial selection, recombination and fitness measure to generate linear p ..."
Abstract - Cited by 18 (0 self) - Add to MetaCart
Abstract. In this paper we describe SINERGY, which is a highly parallelizable, linear planning system that is based on the genetic programming paradigm. Rather than reasoning about the world it is planning for, SINERGY uses artificial selection, recombination and fitness measure to generate linear plans that solve conjunctive goals. We ran SINERGY on several domains (e.g., the briefcase problem and a few variants of the robot navigation problem), and the experimental results show that our planner is capable of handling problem instances that are one to two orders of magnitude larger than the ones solved by UCPOP. In order to facilitate the search reduction and to enhance the expressive power of SINERGY, we also propose two major extensions to our planning system: a formalism for using hierarchical planning operators, and a framework for planning in dynamic environments. 1

Experimental Study of Multipopulation Parallel Genetic Programming.

by F. Fernández, M. Tomassini, W.F. Punch, III, Punch Iii, J.M. Sánchez - GENETIC PROGRAMMING, PROCEEDINGS OF EUROGP'2000 , 2000
"... . The parallel execution of several populations in Evolutionary Algorithms has usually given good results. Nevertheless, researchers have to date drawn conflicting conclusions when using some of the Parallel Genetic Programming models. One aspect of the conflict is population size, since publish ..."
Abstract - Cited by 16 (1 self) - Add to MetaCart
. The parallel execution of several populations in Evolutionary Algorithms has usually given good results. Nevertheless, researchers have to date drawn conflicting conclusions when using some of the Parallel Genetic Programming models. One aspect of the conflict is population size, since published GP works do not agree about whether to use large or small populations. This paper presents an experimental study of a number of common GP test problems. Via our experiments, we discovered that an optimal range of values exists. This assists us in our choice of population size and in the selection of an appropriate Parallel Genetic Programming model. Finding efficient parameters helps us to speed up our search for solutions. At the same time, it allows us to locate features that are common to Parallel Genetic Programming and the classic Genetic Programming Technique. 1

Linear-tree GP and its comparison with other GP structures

by Wolfgang Kantschik, Wolfgang Banzhaf - Genetic Programming, Proceedings of EuroGP’2001, volume 2038 of LNCS , 2001
"... Abstract. In recent years different genetic programming (GP) structures have emerged. Today, the basic forms of representation for genetic programs are tree, linear and graph structures. In this contribution we introduce a new kind of GP structure which we call Linear-tree. We describe the linear-tr ..."
Abstract - Cited by 16 (5 self) - Add to MetaCart
Abstract. In recent years different genetic programming (GP) structures have emerged. Today, the basic forms of representation for genetic programs are tree, linear and graph structures. In this contribution we introduce a new kind of GP structure which we call Linear-tree. We describe the linear-tree-structure, as well as crossover and mutation for this new GP structure in detail. We compare linear-tree programs with linear and tree programs by analyzing their structure and results on different test problems. 1 Introduction of Linear-Tree GP The representations of programs used in Genetic Programming can be classified by their underlying structure into three major groups: (1) tree-based [Koz92,Koz94], (2) linear [Nor94,BNKF98], and (3) graph-based [TV96] representations. This paper introduces a new representation for GP programs. This new representation,

Constructive Induction using Genetic Programming

by Hilan Bensusan, Ibrahim Kuscu - In Evolutionary Computing and Machine Learning Workshop (ICML-96
"... A constructive induction model using genetic programming is presented. The model evolves new attributes starting from a random population of possible attributes constructed as functions of the original attributes. The model is tested on hard supervised learning problems and its performance is compar ..."
Abstract - Cited by 12 (1 self) - Add to MetaCart
A constructive induction model using genetic programming is presented. The model evolves new attributes starting from a random population of possible attributes constructed as functions of the original attributes. The model is tested on hard supervised learning problems and its performance is compared with backpropagation and C4.5. The performance of the system on learning incomplete 4-bit parity is reported to be better. 1 INTRODUCTION Constructive induction (CI) is an effort to improve the attribute vector of a learning problem in order to make the problem more easily learned for a particular learning algorithm(see [ 9 ] , [ 11 ] , [ 21 ] ). CI is often used to tackle hard problems for a given learning algorithm L - problems are hard for L if the training set contains all the relevant information for the induction of the target function but this information cannot be extracted by L (see [ 13 ] ). In general, CI is used to deal with problems that are hard for the most used learning ...

Evolving heuristics for planning

by Ricardo Aler, Daniel Borrajo, Pedro Isasi - Lecture Notes in Computer Science , 1998
"... Abstract. In this paper we describe EvoCK, a new approach to the application of genetic programming (GP) to planning. This approach starts with a traditional AI planner (PRODIGY) and uses GP to acquire control rules to improve its efficiency. We also analyze two ways to introduce domain knowledge ac ..."
Abstract - Cited by 12 (7 self) - Add to MetaCart
Abstract. In this paper we describe EvoCK, a new approach to the application of genetic programming (GP) to planning. This approach starts with a traditional AI planner (PRODIGY) and uses GP to acquire control rules to improve its efficiency. We also analyze two ways to introduce domain knowledge acquired by another method (HAMLET) into EvoCK: seeding the initial population and using a new operator (knowledge-based crossover). This operator combines genetic material from both an evolving population and a non-evolving population containing background knowledge. We tested these ideas in the blocksworld domain and obtained excellent results. 1
(Show Context)

A Scalable Cellular Implementation of Parallel Genetic Programming

by Gianluigi Folino, Clara Pizzuti, Omenico Spezzano - IEEE Transactions on Evolutionary Computation , 2003
"... A new parallel implementation of genetic programming based on the cellular model is presented and compared with both canonical genetic programming and the island model approach. The method adopts a load balancing policy that avoids the unequal utilization of the processors. Experimental results on b ..."
Abstract - Cited by 12 (6 self) - Add to MetaCart
A new parallel implementation of genetic programming based on the cellular model is presented and compared with both canonical genetic programming and the island model approach. The method adopts a load balancing policy that avoids the unequal utilization of the processors. Experimental results on benchmark problems of different complexity show the superiority of the cellular approach with respect to the canonical sequential implementation and the island model. A theoretical performance analysis reveals the high scalability of the implementation realized and allows to predict the size of the population when the number of processors and their efficiency are fixed.

Lineargraph gp - a new gp structure

by Wolfgang Kantschik, Wolfgang Banzhaf - Genetic Programming, Proceedings of EuroGP'2002, LNCS , 2002
"... Abstract. In recent years different genetic programming (GP) structures have emerged. Today, the basic forms of representation for genetic programs are tree, linear and graphstructures. In this contribution we introduce a new kind of GP structure which we call linear-graph. This is a further develop ..."
Abstract - Cited by 11 (2 self) - Add to MetaCart
Abstract. In recent years different genetic programming (GP) structures have emerged. Today, the basic forms of representation for genetic programs are tree, linear and graphstructures. In this contribution we introduce a new kind of GP structure which we call linear-graph. This is a further development to the linear-tree structure that we developed earlier. We describe the linear-graph structure, as well as crossover and mutation for this new GP structure in detail. We compare linear-graph programs withlinear and tree programs by analyzing their structure and results on different test problems. 1 Introduction of Linear-Graph GP This paper introduces a new representation for GP programs. This new representation, named linear-graph, has been developed with the goal of giving a program the flexibility to choose different execution paths for different inputs. The hope is to create programs of higher complexity, so that we can evolve programs
(Show Context)

Improving the Evolvability of Digital Multipliers Using Embedded Cartesian Genetic Programming and Product Reduction

by James Alfred Walker, Julian Francis Miller - Proceedings of 6th International Conference in Evolvable Systems. Springer, LNCS 3637 , 2005
"... Abstract. Embedded Cartesian Genetic Programming (ECGP) is a form of Ge-netic Programming based on an acyclic directed graph representation. In this paper we investigate the use of ECGP together with a technique called Product Reduction (PR) to reduce the time required to evolve a digital multiplier ..."
Abstract - Cited by 10 (5 self) - Add to MetaCart
Abstract. Embedded Cartesian Genetic Programming (ECGP) is a form of Ge-netic Programming based on an acyclic directed graph representation. In this paper we investigate the use of ECGP together with a technique called Product Reduction (PR) to reduce the time required to evolve a digital multiplier. The results are compared with Cartesian Genetic Programming (CGP) with and without PR and show that ECGP improves evolvability and also that PR im-proves the performance of both techniques by up to eight times on the digital multiplier problems tested. 1
(Show Context)

Synthesis of low-sensitivity second-order digital filters using genetic programming with automatically defined functions

by Kazuyoshi Uesaka, Masayuki Kawamata - IEEE Signal Processing Letters , 2000
"... This paper proposes a synthesis method for low coefficient sensitivity second-order IIR digital filter structures using Genetic Programming with Automatically Defined Func-tions (GP-ADF). In this paper, digital filter structures are represented as S-expressions with subroutines. It is easy to genera ..."
Abstract - Cited by 9 (1 self) - Add to MetaCart
This paper proposes a synthesis method for low coefficient sensitivity second-order IIR digital filter structures using Genetic Programming with Automatically Defined Func-tions (GP-ADF). In this paper, digital filter structures are represented as S-expressions with subroutines. It is easy to generate syntactically valid S-expressions and perform the genetic operations because the representation is suitable for GP. In a numerical example, we use the fitness measure in-cluding the magnitude sensitivity, and demonstrate that the proposed method can synthesize efficiently very low coeffi-cient sensitivity filter structures. 1.
(Show Context)