Browsing by Author "Chire Saire, Josimar Edinson"

Now showing 1 - 4 of 4
  • No Thumbnail Available
    Item
    An approach to real-coded quantum inspired evolutionary algorithm using particles filter
    (Institute of Electrical and Electronics Engineers Inc., 2016) Chire Saire, Josimar Edinson; Túpac Valdivia, Yván Jesús
    This work proposes, implements and evaluates the FP-QIEA-R model as a new quantum inspired evolutionary algorithm based on the concept of quantum superposition that allows the optimization process to be carried on with a smaller number of evaluations. This model is based on a QIEA-R, but instead of just using quantum individuals based on uniform probability density functions, where the update consists on change the width and mean of each pdf; this proposal uses a combined mechanism inspired in particle filter and multilinear regression, re-sampling and relative frequency with the QIEA-R to estimate the probability density functions in a better way. To evaluate this proposal, some experiments under benchmark functions are presented. The obtained statistics from the outcomes show the improved performance of this proposal optimizing numerical problems. © 2015 IEEE.
  • No Thumbnail Available
    Item
    FP-AK-QIEAR-R in protein folding application
    (Institute of Electrical and Electronics Engineers Inc., 2017) Chire Saire, Josimar Edinson; Túpac Valdivia, Yván Jesús
    There are many Evolutionary Algorithms which main features are: population, evolutionary operations (crossover, mate, mutation and others). Most of them are based on randomness and follow a criteria using fitness like selector. The proposal uses probability density function according to best of initial population to sample new population and save better individuals iteratively. Then using centroid criteria sample for every dimension and get better individuals. It had good results with benchmark functions. A real application was performed with experiments in protein folding and it showed good results. © 2016 IEEE.
  • No Thumbnail Available
    Item
    Parameters analysis of QIEA-R in convergence quality
    (Institute of Electrical and Electronics Engineers Inc., 2017) Chire Saire, Josimar Edinson; Túpac Valdivia, Yván Jesús
    QIEA-R (Quantum Inspired Evolutionary Algorithm with Real Codification) was proposed for solving numerical problems obtaining better results when compared with traditional EAs, DE and PSO algorithms. It is inspired on the concept of quantum superposition in order to reduce the number of evaluations. QIEA-R has two important steps: initialization of the quantum population and updating of the quantum population. This paper analyzes these two steps and parameters related: Size of classical population, number of iterations, over some benchmark functions using statistical measurements to evaluate their importance and effect in convergence quality. The results shows the importance of quantum population size and update frequency. © 2016 IEEE.
  • Thumbnail Image
    Item
    Una propuesta de algoritmo evolutivo de inspiración cuántica para representación real usando filtro de partículas
    (Universidad Católica San Pablo, 2017) Chire Saire, Josimar Edinson; Túpac Valdivia, Yván Jesús
    En este trabajo se propone, implementa y evalúa el modelo Quantum Inspired Evolutionary Algorithm with Real Representation using Filter Particle (FP-QIEA-R); este modelo usa la generación clásica del modelo Quantum Inspired Evolutionary Algorithm with Real Representation (QIEA-R) (uso de función de distribución de probabilidad uniforme) y propone la generación clásica usando un mecanismo inspirado en filtro de partículas, aproximación de funciones, recompensa de los mejores individuos y muestreo usando funciones de distribución de probabilidad para la búsqueda global y centroides para la búsqueda local. Durante el progreso de este trabajo fueron evaluados varios métodos de estimación de funciones: unidimensionales (splines, interpolación de akima), multidimensionales (regresión multilineal, parzen window) para estimar la función de distribución acumulada(modificada usando el criterio de recompensa). Para evaluar el modelo, se realizaron experimentos con funciones benchmark (Ackley, Rastrigin, Rosenbrock, Schwefel, Sphere) usando una dimensionalidad de 30 y 100. Algunas aplicaciones reales fueron evaluadas: la inicialización de una red perceptrón multicapa para ayudar la convergencia(reducir el número de épocas), encontrar los ángulos en el problema de desdoblamiento de proteínas. En los primeros experimentos, todos los modelos fueron comparados usando medidas estadísticas(media,desviación estándar), tiempo de ejecución y de acuerdo a los resultados obtenidos el modelo más robusto fue el modelo que usa interpolación de akima y añade durante las generaciones a los mejores individuos. Los resultados obtenidos mostraron que la propuesta tiene el mejor desempeño tratando diversos problemas de optimización numérica comparado con el modelo existente QIEA-R.