Scheduled System Maintenance:
Some services will be unavailable Sunday, March 29th through Monday, March 30th. We apologize for the inconvenience.
By Topic

Evolutionary Computation, 1994. IEEE World Congress on Computational Intelligence., Proceedings of the First IEEE Conference on

Date 27-29 June 1994

Filter Results

Displaying Results 1 - 25 of 159
  • Evolving better representations through selective genome growth

    Publication Year: 1994 , Page(s): 182 - 187 vol.1
    Cited by:  Papers (5)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (608 KB)  

    The choice of how to represent the search space for a genetic algorithm (GA) is critical to the GA's performance. Representations are usually engineered by hand and fixed for the duration of the GA run. Here a new method is described in which the degrees of freedom of the representation-i.e. the genes-are increased incrementally. The phenotypic effects of the new genes are randomly drawn from a space of different functional effects. Only those genes that initially increase fitness are kept. The genotype-phenotype map that results from this selection during the construction of the genome allows better adaptation. This effect is illustrated with the NK landscape model. The resulting genotype-phenotype maps are much less epistatic than unselected maps would be, having extremely low values of “K”-the number of fitness components affected by each gene. Moreover, these maps are exquisitely tuned to the specifics of the epistatic fitness function, creating adaptive landscapes that are much smoother than generic NK landscapes with the same genotype-phenotype maps, with fitness peaks many standard deviations higher. Thus a caveat should be made when making arguments about the applicability of generic properties of complex systems to evolved systems. This method may help to solve the problem of choice of representations in genetic algorithms. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Proceedings of the First IEEE Conference on Evolutionary Computation. IEEE World Congress on Computational Intelligence

    Publication Year: 1994
    Save to Project icon | Request Permissions | PDF file iconPDF (33 KB)  
    Freely Available from IEEE
  • A combined neural and genetic learning algorithm

    Publication Year: 1994 , Page(s): 770 - 774 vol.2
    Cited by:  Papers (4)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (312 KB)  

    Neural networks and genetic algorithms are well-known representatives of learning procedures. In this paper a hybrid procedure, which combines both concepts, is introduced. Its functionality is presented on a typical pattern recognition problem View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Evolutionary generation and training of recurrent artificial neural networks

    Publication Year: 1994 , Page(s): 759 - 763 vol.2
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (344 KB)  

    An evolutionary artificial neural network training and design methodology is presented, aimed at obtaining optimum or quasi-optimum synchronous recurrent neural networks capable of processing sequential inputs. We show that, through the use of this method and working with floating point and integer valued chromosomes, it is possible to achieve optimum results, considering very small populations and few generations. In order to implement this methodology, we have developed GENIAL, a genetic algorithm development environment which is specifically designed for solving this type of problem. It offers ways of testing adequate fitness functions and many tools for improving results. Finally, we comment on the sequential introduction of different constraints in genetic algorithms, presenting a classical example where several design requirements are met simultaneously and which demonstrates the power of this method View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Finding improved simulated annealing schedules with genetic programming

    Publication Year: 1994 , Page(s): 391 - 395 vol.1
    Cited by:  Papers (1)  |  Patents (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (416 KB)  

    Many combinatorial problems are too difficult to be solved optimally, and hence heuristics are used to obtain “good” solutions in “reasonable” time. A heuristic that has been successfully applied to a variety of problems is simulated annealing. However, the performance of simulated annealing strongly depends on the appropriate choice of a key parameter, the annealing schedule. Usually, researchers experiment with a number of manually created annealing schedules and then choose the one that performs best for their algorithms. This work applies genetic programming to replace this manual search. For a given problem, we search for an optimal annealing schedule. We demonstrate the potential of this new approach by optimizing the annealing schedule for one of the hardest combinatorial optimization problem, the quadratic assignment problem. We introduce a new algorithm for solving the quadratic assignment problem that performs extremely well, and we outline properties of good annealing schedules View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neuro-genetic truck backer-upper controller

    Publication Year: 1994 , Page(s): 720 - 723 vol.2
    Cited by:  Papers (6)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (324 KB)  

    The precise docking of a truck at a loading dock has been proposed in (Nguyen and Widrow, 1990) as a benchmark problem for non-linear control by neural-nets. The main difficulty is that backpropagation is not a priori suitable as a learning paradigm, because no set of training vectors is available: It is non-trivial to find solution trajectories that dock the truck from anywhere in the loading yard. In this paper we show how a genetic algorithm can evolve the weights of a feedforward 3-layer neural net that solves the control problem for a given starting state, achieving a short trajectory from starting point to goal. The fitness of a net in the population is a function of both the nearest position from the goal and the distance travelled. The influence of input data renormalisation on trajectory precision is also discussed View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Applying crossover operators to automatic neural network construction

    Publication Year: 1994 , Page(s): 750 - 752, 752a vol.2
    Cited by:  Papers (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (304 KB)  

    The ability to automatically construct neural networks is of importance, since it supports reduction in development time and can lead to simpler designs than traditionally handcrafted networks. Automation is further required to take the step towards a more autonomous learning system. In this paper, we report further results involving the automatic network construction algorithm EGP (Evolutionary Growth Perceptron), which utilizes simple evolutionary processes to locally train network features using the perceptron rule. Emphasis is placed on determining the effectiveness of several types of crossover operators in conjunction with varying the population size and the number of epochs during which individual perceptrons are trained. The crossover operators considered and introduced are: simple random, weighted and blocked View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Dynamic scheduling of computer tasks using genetic algorithms

    Publication Year: 1994 , Page(s): 829 - 833 vol.2
    Cited by:  Papers (4)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (404 KB)  

    We concentrate on non-preemptive hard real-time scheduling algorithms. We compare FIFO, EDLF, SRTF and genetic algorithms for solving this problem. The objective of the scheduling algorithm is to dynamically schedule as many tasks as possible such that each task meets its execution deadline, while minimizing the total delay time of all of the tasks. We present a MicroGA that uses a small population size of 10 chromosomes, running for 10 trials using a rather high mutation rate with a sliding window of 10 tasks. The steady-state GA was determined to be better than the generational GA for our MicroGA. We also present a parallel MicroGA model designed for parallel processors. The parallel MicroGA works best when migration is used to move tasks from one processor to another to even out the load as much a possible. Test cases show that the sequential MicroGA model and the parallel MicroGA model produced superior task schedules compared to other algorithms tested View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A comparison of simulated evolution and genetic evolution performance

    Publication Year: 1994 , Page(s): 374 - 378 vol.1
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (300 KB)  

    Simulated evolution (SE) and the genetic algorithm (GA) are closely related methods for finding optimal solutions by directed random search. Both methods start with a population of randomly selected trial solutions and use that information to “evolve” a next generation of trials which, on the average, has improved fitness (i.e., is closer to the optimum). As the population average improves, so too, is the best of the population likely to be closer to the optimal solution sought. The methods differ principally in the manner by which each new generation evolves from the previous generation. In addition, the GA represents the search space by discrete points so that the optimum is limited to one of those points. The SE method treats the search space as continuous. The paper makes a side-by-side comparison of the performance of the two methods when applied to the same problem. The problem to be solved is sometimes called the Dictator Problem. It is to maximize the minimum angular separation, α among n points placed on a sphere View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Evolutionary design of FIR digital filters with power-of-two coefficients

    Publication Year: 1994 , Page(s): 110 - 114 vol.1
    Cited by:  Papers (2)  |  Patents (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (312 KB)  

    The paper presents a genetic approach to the design of finite impulse response filters with coefficients constrained to be sums of power-of-two terms. The evolutionary algorithm is explained and compared experimentally with other state-of-the-art design methods. The proposed technique is able to attain good results and can be easily implemented on parallel hardware View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Elitist recombination: an integrated selection recombination GA

    Publication Year: 1994 , Page(s): 508 - 512 vol.1
    Cited by:  Papers (18)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (400 KB)  

    In standard genetic algorithms selection and recombination are separated into two distinct phases. Here we discuss an extremely simple GA implementation where selection and recombination are intertwined. Competition for survival takes place at the level of each family-the mating parents and their offspring-which results in a local elitist selection operator. We derive an analytical model for optimizing the bit counting function, and compare the elitist recombination GA with tournament selection and standard recombination. Results suggest that elitist recombination is less sensitive to undersized populations, while there is no need to choose a specific value for the crossover probability View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Genetic algorithm and simulated annealing for optimal robot arm PID control

    Publication Year: 1994 , Page(s): 707 - 713 vol.2
    Cited by:  Papers (15)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (420 KB)  

    This paper describes the use of genetic algorithm (GA) and simulated annealing (SA) for optimizing the parameters of PID controllers for a 6-DOF robot arm. A GA and a SA are designed to optimal-tune the parameters of the PID controller of each joint for a single step response and for the tracking of other specified trajectories. The GA and the SA are required to optimize evaluation functions related to the combinations of different performance indices. Simulations are carried out on a PUMA 560 arm model being controlled by PID controllers with their parameters optimized using the proposed GA and SA. Based on the simulation results, the performances of genetic algorithm and simulated annealing are compared and discussed View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An anti-adaptationist approach to genetic algorithms

    Publication Year: 1994 , Page(s): 619 - 623 vol.2
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (344 KB)  

    Most current implementations of the Genetic Algorithm (GA) closely follow an Adaptationist theory of evolutionary biology. This approach impedes finding a solution or leads to the reduction of robustness in solutions. In this paper, we show how an Anti-Adaptationist approach to GAs would more closely follow biological example, and aid in finding satisfactory solutions to systems that must operate in dynamic environments, or that have many complex and interdependent features. We then explain necessary modifications to the GA to implement these changes View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An empirical comparison of two evolutionary methods for satisfiability problems

    Publication Year: 1994 , Page(s): 450 - 455 vol.1
    Cited by:  Papers (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (484 KB)  

    The paper compares two evolutionary methods for model finding in the satisfiability problem (SAT): genetic algorithms (GAs) and the mask method (MASK). The main characteristics of these two methods are that both of them are population-based, and use binary representation. Great care is taken to make sure that the same SAT instances and the same criteria are used in the comparison. Results indicate that MASK greatly outperforms GAs in the sense that MASK manages to deal with harder SAT instances at a lower cost View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Automated feature extraction for supervised learning

    Publication Year: 1994 , Page(s): 674 - 679 vol.2
    Cited by:  Patents (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (444 KB)  

    Feature extraction has traditionally been a manual process and something of an art. Methods derived from statistics and linear systems theory have been proposed, but by general consensus effective feature extraction remains a difficult problem. Recently W. Tackett (1993) showed that genetic programming (GP) can be effective in automatically constructing features for identifying potential targets in digital images with high accuracy. From a basis set of simple arithmetic functions, he was able to construct numerical features that outperformed manually-constructed features when used as inputs to several classifiers, including a binary-tree classifier and a multi-layer perceptron trained by back-propagation. Seeking a more generic feature-construction procedure, we developed a GP-based algorithm to extract features in a variety of domains and for most classification methods, including decision trees, feed-forward neural networks, and Bayesian classifiers. We have tested the technique with success by extracting features for three different types of problems: Boolean functions with binary features, a NASA telemetry problem with multiple classes and real-valued time-series inputs, and a wine variety classification problem with real-valued features from the UCI Machine Learning repository. We formally define the feature-construction method and show in some detail how it applies to specific classification problems View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Selective pressure in evolutionary algorithms: a characterization of selection mechanisms

    Publication Year: 1994 , Page(s): 57 - 62 vol.1
    Cited by:  Papers (36)  |  Patents (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (584 KB)  

    Due to its independence of the actual search space and its impact on the exploration-exploitation tradeoff, selection is an important operator in any kind of evolutionary algorithm. All important selection operators are discussed and quantitatively compared with respect to their selective pressure. The comparison clarifies that only a few really different and useful selection operators exist: proportional selection (in combination with a scaling method), linear ranking, tournament selection, and (μ,λ)-selection (respectively (μ+λ)-selection). Their selective pressure increases in the order as they are listed here. The theoretical results are confirmed by an experimental investigation using a genetic algorithm with different selection methods on a simple unimodal objective function View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Knowledge-based nonuniform crossover

    Publication Year: 1994 , Page(s): 22 - 27 vol.1
    Cited by:  Papers (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (288 KB)  

    One-point, two-point and k-point crossover can be viewed as special cases of uniform crossover, where genetic material is chosen each locus of either parent with equal probability (G. Syswerda, 1989). The paper generalizes uniform crossover to “non-uniform crossover” using “mask” vectors whose elements are real numbers ∈[0, 1], representing problem-specific knowledge that improves performance by biasing the selection of alleles from either parent. This knowledge based non-uniform crossover (KNUX) is applied to two NP optimization problems: graph partitioning and soft-decision decoding of linear block codes (H.S. Maini, 1993). Simulation results show orders of magnitude improvement of this operator over two-point and uniform crossover. An appropriate schema theorem is also developed View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • GAVaPS-a genetic algorithm with varying population size

    Publication Year: 1994 , Page(s): 73 - 78 vol.1
    Cited by:  Papers (48)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (468 KB)  

    The size of the population can be critical in many applications of genetic algorithms. If the population size is too small, the genetic algorithm may converge too quickly; if it is too large, the genetic algorithm may waste computational resources; the waiting time for an improvement might be too long. We propose an adaptive method for maintaining variable population size, which grows and shrinks together according to some characteristic of the search. The first experimental results indicate some merits of the proposed method View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Bach in a box: the evolution of four part Baroque harmony using the genetic algorithm

    Publication Year: 1994 , Page(s): 852 - 857 vol.2
    Cited by:  Papers (10)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (508 KB)  

    The space of all possible musical compositions given the constraints of the standard Western tonal and rhythmic system is uncountably vast. We look at a small subsection of this space: Baroque harmony. The rules governing Baroque harmony have been carefully laid out by musical scholars, and the size of this search space becomes tractable if we are given a melody and a key signature to work with. Music is easily represented in numerical form, and is thus an obvious candidate for computer manipulation. We investigate a system using the genetic algorithm that is capable of generating well constructed Baroque-style harmonies given a user-defined melody View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Improving classification performance in the bumptree network by optimising topology with a genetic algorithm

    Publication Year: 1994 , Page(s): 490 - 495 vol.1
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (448 KB)  

    This paper presents a successful synthesis of evolutionary and connectionist methods, based on the genetic optimisation of a recently introduced neural network model, the bumptree network. We show that the bumptree network is inherently more suited to optimisation by a genetic algorithm (GA) than other neural network models such as the multi-layer perceptron (MLP). We describe a hierarchical genetic coding which addresses the problem of representing certain strong dependencies which exist between the bumptree's structural parameters, and show that our coding scheme has the desirable properties of continuity, isomorphism, completeness, closure and low redundancy with respect to the space of possible bumptree structures. We present empirical results which show that bumptree networks evolved by the GA significantly outperform the orthodox bumptree on several tasks, including the difficult real-world classification task of spoken vowel recognition View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An enhanced operator-oriented genetic search algorithm

    Publication Year: 1994 , Page(s): 235 - 238 vol.1
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (292 KB)  

    This paper proposes a new search process incorporated into an operator-oriented genetic algorithm (GA). The new search algorithm solves problems in the context of invertible symbolic operations on a combinational finite state environment. The algorithm exploits the GA's ability to search for solutions without regard to a priori knowledge of the problem domain. The validity of the algorithm is illustrated by solving Rubik's cube View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A delayed-action classifier system for learning in temporal environments

    Publication Year: 1994 , Page(s): 670 - 673 vol.2
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (292 KB)  

    This paper describes a modified version of the traditional classifier system called the Delayed Action Classifier System (DACS) which has been conceived for learning in environments that exhibit a rich temporal structure. DACS operates by delaying the action of appropriately tagged classifiers (called `delayed-action classifiers') by a number of execution cycles which is encoded on the action part of these classifiers. This modification allows the rule discovery strategy, in many instances a genetic algorithm, to simultaneously explore the spaces of action (what to do) and time (when to do it). Results of initial experiments, which appear encouraging, of applying DACS to a prediction problem are presented, and the possible application of the delayed-action idea to learning in real-time environments is discussed View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Improving a vehicle routing heuristic through genetic search

    Publication Year: 1994 , Page(s): 194 - 199 vol.1
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (324 KB)  

    A genetic algorithm is applied to the search of good parameter settings for a vehicle routing heuristic. The parameter settings identified by the genetic search allow the insertion heuristic to generate solutions that are much better than the solutions previously reported on a standard set of routing problems View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • VLSI circuit synthesis using a parallel genetic algorithm

    Publication Year: 1994 , Page(s): 104 - 109 vol.1
    Cited by:  Papers (6)  |  Patents (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (420 KB)  

    A parallel implementation of a genetic algorithm used to evolve simple analog VLSI circuits is described. The parallel computer system consisted of twenty distributed SPARC workstations whose computational activity is controlled by the parallel environment coordination language Linda. Work-in-progress on using the parallel GA to realize optimized circuits and to discover new types of equivalent-function circuits is presented. The use of biologically inspired development rules to limit the scope of circuits generated by recombination operators to circuits that have an increased chance of surviving is briefly discussed View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A parallel genetic algorithm on the CM-2 for multi-modal optimization

    Publication Year: 1994 , Page(s): 818 - 822 vol.2
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (356 KB)  

    A genetic algorithm is an optimization method well suited to be implemented on a SIMD machine; it deals with a population of individuals (Multiple Data) that evolve in parallel and undergo the same operations (Single Instruction). This paper presents a genetic algorithm with a dynamic division mechanism conceived on the Connection Machine-2 to treat multimodal optimization problems, i.e. search spaces with multiple maxima. The general idea of the algorithm is to dynamically divide the population into an increasing number of subpopulations to allow specialization on the different maxima discovered during the search process. The method is flexible because it requires practically no a-priori knowledge about the fitness function. Results of applications to multi-modal two-dimensional landscapes are presented View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.