This work provides an analysis of using the evolutionary algorithm EPNet to create ensembles of artificial neural networks to solve a range of forecasting tasks. Several previous studies have tested the EPNet algorithm in the classification field, taking the best individuals to solve the problem and creating ensembles to improve the performance. But no studies have analyzed the behavior of the algorithm in detail for time series forecasting, nor used ensembles to try to improve the predictions. Thus, the aim of this work is to compare the ensemble approach, using two linear combination methods to calculate the output, against the best individual found. Since there are several parameters to adjust, experiments are set up to optimize them and improve the performance of the algorithm. The algorithm is tested on 21 time series of different behaviors. The experimental results show that, for time series forecasting, it is possible to improve the performance by using the ensemble method rather than using the best individual. This demonstrates that the information contained in the EPNet population is better than the information carried by any one individual.

No abstract

Resumen. En el presente trabajo se utilizó la segmentación de letras de fotografías tomadas mediante la cámara del Parrot AR Drone, con el objetivo de establecer una interacción de estímulo-respuesta, donde la imagen original en formato de combinación de colores Rojo, Verde y Azul (RGB) se segmentó por color (escogiendo el canal rojo). Una vez que se reconoce el carácter, el Drone ejecuta la acción correspondiente. Se utilizaron inicialmente patrones numéricos libres de ruido y posteriormente se agregaron algunos pixeles a la imagen con el objetivo de hacer más robusto este conjunto de patrones, los cuales proporcionaron el conjunto de entrenamiento para la red neuronal y de esta forma se pudieron interpolar patrones nuevos. Para la segmentación de imágenes se utilizaron técnicas de detección de bordes que incluyen el filtro de Sobel así como filtros para eliminación de ruido basados en filtrado de la mediana que es un filtro pasa baja. Todo lo anterior se llevó a cabo en un entorno cerrado y se espera ampliar este trabajo para su aplicación en diferentes entornos.Palabras clave: Segmentación de caracteres, vehículos aéreos no tripulados (UAVs), procesamiento de imágenes, filtro Sobel, detección de bordes, redes neuronales artificiales.Abstract. In this paper the letter segmentation of photographs was used, taken from a Parrot AR Drone's camera with the aim of establishing a stimulusresponse, where the original picture formed by Red, Green and Blue (RGB) colors was segmented by color (choosing the red channel). Once the character is recognized, the Drone executes the corresponding action. Noise-free number patterns were initially used and then some pixels were added in the image in order to make a set of patterns more robust, which provided the training set for neural network and thus are able to interpolate new patterns. Edge techniques detection were used for image segmentation including Sobel filter and filters for noise removal based on the median filtering, that is a low pass filter. All this took place in a closed environment, expecting to extend this to different environments.

In recent years, Evolutionary Algorithms (EAs) have been remarkably useful to improve the robustness of Artificial Neural Networks (ANNs). This study introduces an experimental analysis using an EAs aimed to evolve ANNs architectures (the FS-EPNet algorithm) to understand how neural networks are evolved with a steady-state algorithm and compare the Singlestep-ahead (SSP) and Multiple-step-ahead (MSP) methods for prediction tasks over two test sets. It was decided to test an inside-set during evolution and an outside-set after the whole evolutionary process has been completed to validate the generalization performance with the same method (SSP or MSP). Thus, the networks may not be correctly evaluated (misleading fitness) if the single SSP is used during evolution (inside-set) and then the MSP at the end of it (outsideset). The results show that the same prediction method should be used in both evaluation sets providing smaller errors on average.

In computer sciences, matrices are widely used for representing different kinds of information. Measuring similarity among square matrices is an interesting open problem in computer sciences. Furthermore, eigenvalues and eigenvectors are a powerful way for representing and characterizing square matrices. In this paper we introduce a new similarity measure among two square data matrices of the same class; the idea is based on evaluating the effect of conjugate the eigenvalues and eigenvectors of one matrix with the other matrix, and vice versa. Some experimental results are showed in order to analyze and exemplify the Eigenconjugation as an approach for the problem of similarity of matrices.