Genetic Algorithm Learning Research Articles

Genetic approach to the design of associative memory

In this paper, a genetic learning algorithm for neuron-weighted associative memory is presented. According to the criterions that measures the stability and attraction of associative memory, we cast the learning procedure into a global constrained maximization problem, solved by a genetic algorithm. This learning rule guarantees the storage of training patterns with basins of attraction as large as possible, which is realized by an appropriately defined discrete evaluation index. A larger attraction basin implies higher noise correction ability of associative memory. However, automatic adjustment of the attraction basin is difficult to realize by other method. To evaluate the performance of our learning strategy, a large number of simulation have been carried out

Adaptive fuzzy genetic learning algorithm for the hybrid position / force control of constrained flexible-link robotic manipulators

In this paper, the hybrid position and force control problem is studied for a class of planar two-link constrained flexible link manipulators. An adaptive fuzzy genetic learning algorithm is proposed to update the parameters of the controller, and to control the tip position and the contact force so as to enhance the control precision and improve the robustness of the systems. Simulation results show that this newly presented control scheme is of very good performances.

Learning and behavioral stability

This article tries to connect two separate strands of literature concerning genetic algorithms. On the one hand, extensive research took place in mathematics and closely related sciences in order to find out more about the properties of genetic algorithms as stochastic processes. On the other hand, recent economic literature uses genetic algorithms as a metaphor for social learning. This paper will face the question of what an economist can learn from the mathematical branch of research, especially concerning the convergence and stability properties of the genetic algorithm. It is shown that genetic algorithm learning is a compound of three different learning schemes. First, each particular scheme is analyzed. Then it is shown that it is the combination of the three schemes that gives genetic algorithm learning its special flair: A kind of stability somewhere in between asymptotic convergence and explosion.

Coevolution and stable adjustments in the cobweb model

The paper is concerned with genetic algorithm learning in a cobweb economy. Besides discussing several specification details in the genetic operators, the model includes four different types of firm forecasting rules and subjects the demand side to serially correlated random shocks. The main finding of the simulation experiments is that the genetic algorithm is a reasonably good approximation of the moving Walrasian equilibria, and that this process is characterized by the coevolution of different strategies. Accordingly, it is just the persistent heterogeneity of firms, and the persistently changing composition of this heterogeneity, that achieves stability. In this world, convergence is improved by weak, rather than strong, evolutionary pressure.

Review of "Genetic Algorithms for Machine Learning by John J. Greffenstette", Kluwer Academic Publishers, 1993.

This book is a reprint from a special issue of the journal Machine Learning on Genetic Algorithms (GAs). The fact that it was the third special issue of the journal on GAs shows the continuous increase of interest of people in the field. The five selected articles were presented at the Fourth International Conference on Genetic Algorithms, in June 1991 (San Diego), and at a Special Workshop for Machine Learning at the same conference. The intended book audience is both researchers and practitioners in the field.

Texture Classification and Segmentation by Cellular Neural Networks Using Genetic Learning

We present a new single-chip texture classifier based on the cellular neural network (CNN) architecture. Exploiting the dynamics of a locally interconnected 2D cell array of CNNs we have developed a theoretically new method for texture classification and segmentation. This technique differs from other convolution-based feature extraction methods since we utilize feedback convolution, and we use a genetic learning algorithm to determine the optimal kernel matrices of the network. The CNN operators we have found for texture recognition may combine different early vision effects. We show how the kernel matrices can be derived from the state equations of the network for convolution/deconvolution and nonlinear effects. The whole process includes histogram equalization of the textured images, filtering with the trained kernel matrices, and decision-making based on average gray-scale or texture energy of the filtered images. We present experimental results using digital CNN simulation with sensitivity analysis for noise, rotation, and scale. We also report a tested application performed on a programmable 22 × 20 CNN chip with optical inputs and an execution time of a few microseconds. We have found that this CNN chip with a simple 3 × 3 CNN kernel can reliably classify four textures. Using more templates for decision-making, we believe that more textures can be separated and adequate texture segmentation (< 1% error) can be achieved.

Evolving Gaussian RBF network for nonlinear timeseries modellingand prediction

A genetic algorithm and recursive least squares (RLS) learning algorithm for a Gaussian radial basis function network is described, for modelling and predicting nonlinear time series. Better generalisation performance can be achieved than that of the usual clustering and RLS method.

Contact Friction Compensation for Robots Using Genetic Learning Algorithms

In this paper, the issues of contact friction compensation for constrained robots are presented. The proposed design consists of two loops. The inner loop is for the inverse dynamics control which linearizes the system by canceling nonlinear dynamics, while the outer loop is for friction compensation. Although various models of friction have been proposed in many engineering applications, frictional force can be modeled by the Coulomb friction plus the viscous force. Based on such a model, an on-line genetic algorithm is proposed to learn the friction coefficients for friction model. The friction compensation control input is also implemented in terms of the friction coefficients to cancel the effect of unknown friction. By the guidance of the fitness function, the genetic learning algorithm searches for the best-fit value in a way like the natural surviving laws. Simulation results demonstrate that the proposed on-line genetic algorithm can achieve good friction compensation even under the conditions of measurement noise and system uncertainty. Moreover, the proposed control scheme is also found to be feasible for friction compensation of friction model with Stribeck effect and position-dependent friction model.

Job shop scheduling with a genetic algorithm and machine learning

Dynamic job shop scheduling has been proven to be an intractable problem for analytical procedures. Recent advances in computing technology, especially in artificial intelligence, have alleviated this problem by intelligently restricting the search space considered, thus opening the possibility of obtaining better results. Researchers have used various techniques that were developed under the general rubric of artificial intelligence to solve job shop scheduling problems. The most common of these have been expert systems, genetic algorithms and machine learning. Of these, we identify machine learning and genetic algorithms to be promising for scheduling applications in a job shop. In this paper, we propose to combine complementarily the strengths of genetic algorithms and induced decision trees, a machine learning technique, to develop a job shop scheduling system. Empirical results, using machine learning for releasing jobs into the shop floor and a genetic algorithm to dispatch jobs at each machine, are promising.

Neural network design with genetic learning for control of a single link flexible manipulator

The application of neural networks for active control of lightly damped systems is considered. The training process of the neural-network controller is based on the genetic learning algorithm. The scheme imitates nature's cleansing phenomena of natural selection and survival of the fittest to generate individual controllers with the best fitness values. It essentially incorporates an exhaustive search in the weight-space governed by the rituals of crossover and mutation to seek the optimum neural-network weights to satisfy certain performance criteria. Several appropriate modifications of the classical genetic algorithm for neural-network control purposes are discussed. The genetic-trained neural-network controller is applied for tip position tracking and vibration suppression of a single-link flexible arm. Simulation studies are presented to validate the effectiveness of the advocated algorithms.

Genetic algorithms for learning in fuzzy relational structures

We propose an approach of integrated genetic learning in construction of fuzzy relational architectures described by fuzzy relational equations. These equations have been widely utilized, e.g., in fuzzy model identification, fuzzy control, and fuzzy controllers. While the theoretical foundations of the equations are well developed, they still call for more efficient and diversified schemes of learning. The paper addresses this issue by synergistically combining some fundamental concepts of genetic algorithms (GA) and standard gradient-based techniques into a unified scheme of a stratified learning. Especially, we reveal how the ideas of GAs can be effectively used at the level of the initialization of the gradient-based learning schemes. It is also elucidated how an optimal subpopulation of the strings can be helpful in forming feasibility regions (guarding zones) indispensable for supervising the subsequent phase of the parametric learning.

Genetic algorithm learning and the cobweb model

This paper presents the cobweb model in which competitive firms, in a market for a single good, use a genetic algorithm to update their decision rules about next-period production and sales. The results of simulations show that the genetic algorithm converges to the rational expectations equilibrium for a wider range of parameter values than other algorithms frequently studied within the context of the cobweb model. Price and quantity patterns generated by the genetic algorithm are also compared to the data of experimental cobweb economies. It is shown that the algorithm can capture several features of the experimental behavior of human subjects better than three other learning algorithms that are considered.

Combining evolution with credit apportionment: A new learning algorithm for neural nets

A new approach is presented to Darwinian learning over a repertoire of competing neural nets that includes a credit apportionment algorithm for individual neurons within the networks. The architecture of the system is organized to allow the output of any given layer of a selected network to be available both to the next layer of this same network and also to the next layer of all of the other networks in the repertoire. The pathway through the selected network is considered the (temporarily) dominant pathway through the system architecture. This provides the basis for implementing a competition among single neurons across all of the networks. The firing behavior of the neurons in many networks is directly compared with the corresponding neurons in the dominant network. Payoffs accorded to neurons in one network can thus be shared with neurons in other networks. Neurons with the highest total payoffs will tend to be those that are most compatible with all of the network contexts associated with the best performance on any of the inputs. The algorithm is tested with connectionist-type units. Results include: successful learning experiments with temporal processing tasks using recurrent networks and several types of static learning tasks using analog input vectors. Performance comparisons are given with a standard genetic algorithm learning approach and, for the static learning problems, with standard back propagation as well.

Using Genetic Algorithms to Model the Evolution of Heterogeneous Beliefs

We study a general equilibrium system where agents have heterogeneous beliefs concerning realizations of possible outcomes. The actual outcomes feed back into beliefs thus creating a complicated nonlinear system. Beliefs are updated via a genetic algorithm learning process which we interpret as representing communication among agents in the economy. We are able to illustrate a simple principle: genetic algorithms can be implemented so that they represent pure learning effects (i.e., beliefs updating based on realizations of endogenous variables in an environment with heterogeneous beliefs). Agents optimally solve their maximization problem at each date given their beliefs at each date. We report the results of a set of computational experiments in which we find that our population of artificial adaptive agents is usually able to coordinate their beliefs so as to achieve the Pareto superior rational expectations equilibrium of the model.

Using genetic algorithms for concept learning

In this article, we explore the use of genetic algorithms (GAs) as a key element in the design and implementation of robust concept learning systems. We describe and evaluate a GA-based system called GABIL that continually learns and refines concept classification rules from its interaction with the environment. The use of GAs is motivated by recent studies showing the effects of various forms of bias built into different concept learning systems, resulting in systems that perform well on certain concept classes (generally, those well matched to the biases) and poorly on others. By incorporating a GA as the underlying adaptive search mechanism, we are able to construct a concept learning system that has a simple, unified architecture with several important features. First, the system is surprisingly robust even with minimal bias. Second, the system can be easily extended to incorporate traditional forms of bias found in other concept learning systems. Finally, the architecture of the system encourages explicit representation of such biases and, as a result, provides for an important additional feature: the ability todynamically adjust system bias. The viability of this approach is illustrated by comparing the performance of GABIL with that of four other more traditional concept learners (AQ14, C4.5, ID5R, and IACL) on a variety of target concepts. We conclude with some observations about the merits of this approach and about possible extensions.

The Ninth International Conference on Machine Learning

The Ninth International Conference on Machine Learning was held in Aberdeen, Scotland, from 1-3 July 1992, with 198 participants in attendance. The conference covered a broad range of topics drawn from the general area of machine learning, including concept-learning algorithms, clustering, speedup learning, formal analysis of learning systems, neural networks, genetic algorithms, and applications of machine learning. This article briefly touches on six selected talks that were of exceptional interest.