Allende, H., Bravo, D., & Canessa, E. (2010). Robust design in multivariate systems using genetic algorithms. Qual. Quant., 44(2), 315–332.
Abstract: This paper presents a methodology based oil genetic algorithms, which finds feasible and reasonably adequate Solutions to problems of robust design in multivariate systems. We use a genetic algorithm to determine the appropriate control factor levels for simultaneously optimizing all of the responses of the system, considering the noise factors which affect it. The algorithm is guided by a desirability function which works with only one fitness function although the system May have many responses. We validated the methodology using data obtained from a real system and also from a process simulator, considering univariate and multivariate systems. In all cases, the methodology delivered feasible solutions, which accomplished the goals of robust design: obtain responses very close to the target values of each of them, and with minimum variability. Regarding the adjustment of the mean of each response to the target value, the algorithm performed very well. However, only in some of the multivariate cases, the algorithm was able to significantly reduce the variability of the responses.
|
Allende, H., Elias, C., & Torres, S. (2004). Estimation of the option prime: Microsimulation of backward stochastic differential equations. Int. Stat. Rev., 72(1), 107–121.
Abstract: A mathematical statistical model is needed to obtain an option prime and create a hedging strategy. With formulas derived from stochastic differential equations, the primes for US Dollar/Chilean Pesos currency options using a prime calculator are obtained. Furthermore, a backward simulation of the option prime trajectory is used with a numerical method created for backward stochastic differential equations. The use of statistics in finance is highly important in order to develop complex products.
|
Allende, H., Salas, R., & Moraga, C. (2003). A robust and effective learning algorithm for feedforward neural networks based on the influence function. Lect. Notes Comput. Sc., 2652, 28–36.
Abstract: The learning process of the Feedforward Artificial Neural Networks relies on the data, though a robustness analysis of the parameter estimates of the model must be done due to the presence of outlying observations in the data. In this paper we seek the robust properties in the parameter estimates in the sense that the influence of aberrant observations or outliers in the estimate is bounded so the neural network is able to model the bulk of data. We also seek a trade off between robustness and efficiency under a Gaussian model. An adaptive learning procedure that seeks both aspects is developed. Finally we show some simulations results applied to the RESEX time series.
|
Allende-Cid, H., Canessa, E., Quezada, A., & Allende, H. (2011). An Improved Fuzzy Rule-Based Automated Trading Agent. Stud. Inform. Control, 20(2), 135–142.
Abstract: In this paper an improved Fuzzy Rule-Based Trading Agent is presented. The proposal consists in adding machine-learning-based methods to improve the overall performance of an automated agent that trades in futures markets. The modified Fuzzy Rule-Based Trading Agent has to decide whether to buy or sell goods, based on the spot and futures time series, gaining a profit from the price speculation. The proposal consists first in changing the membership functions of the fuzzy inference model (Gaussian and Sigmoidal, instead of triangular and trapezoidal). Then using the NFAR (Neuro-Fuzzy Autoregressive) model the relevant lags of the time series are detected, and finally a fuzzy inference system (Self-Organizing Neuro-Fuzzy Inference System) is implemented to aid the decision making process of the agent. Experimental results demonstrate that with the addition of these techniques, the improved agent considerably outperforms the original one.
|
Canessa, E., Droop, C., & Allende, H. (2012). An improved genetic algorithm for robust design in multivariate systems. Qual. Quant., 46(2), 665–678.
Abstract: In a previous article, we presented a genetic algorithm (GA), which finds solutions to problems of robust design in multivariate systems. Based on that GA, we developed a new GA that uses a new desirability function, based on the aggregation of the observed variance of the responses and the squared deviation between the mean of each response and its corresponding target value. Additionally, we also changed the crossover operator from a one-point to a uniform one. We used three different case studies to evaluate the performance of the new GA and also to compare it with the original one. The first case study involved using data from a univariate real system, and the other two employed data obtained from multivariate process simulators. In each of the case studies, the new GA delivered good solutions, which simultaneously adjusted the mean of each response to its corresponding target value. This performance was similar to the one of the original GA. Regarding variability reduction, the new GA worked much better than the original one. In all the case studies, the new GA delivered solutions that simultaneously decreased the standard deviation of each response to almost the minimum possible value. Thus, we conclude that the new GA performs better than the original one, especially regarding variance reduction, which was the main problem exhibited by the original GA.
|
Canessa, E., Vera, S., & Allende, H. (2012). A new method for estimating missing values for a genetic algorithm used in robust design. Eng. Optimiz., 44(7), 787–800.
Abstract: This article presents an improved genetic algorithm (GA), which finds solutions to problems of robust design in multivariate systems with many control and noise factors. Since some values of responses of the system might not have been obtained from the robust design experiment, but may be needed in the search process, the GA uses response surface methodology (RSM) to estimate those values. In all test cases, the GA delivered solutions that adequately adjusted the mean of the responses to their corresponding target values and with low variability. The GA found more solutions than the previous versions of the GA, which makes it easier to find a solution that may meet the trade-off among variance reduction, mean adjustment and economic considerations. Moreover, RSM is a good method for estimating the mean and variance of the outputs of highly non-linear systems, which makes the new GA appropriate for optimizing such systems.
|
Fernandez, C., Valle, C., Saravia, F., & Allende, H. (2012). Behavior analysis of neural network ensemble algorithm on a virtual machine cluster. Neural Comput. Appl., 21(3), 535–542.
Abstract: Ensemble learning has gained considerable attention in different learning tasks including regression, classification, and clustering problems. One of the drawbacks of the ensemble is the high computational cost of training stages. Resampling local negative correlation (RLNC) is a technique that combines two well-known methods to generate ensemble diversity-resampling and error negative correlation-and a fine-grain parallel approach that allows us to achieve a satisfactory balance between accuracy and efficiency. In this paper, we introduce a structure of the virtual machine aimed to test diverse selection strategies of parameters in neural ensemble designs, such as RLNC. We assess the parallel performance of this approach on a virtual machine cluster based on the full virtualization paradigm, using speedup and efficiency as performance metrics, for different numbers of processors and training data sizes.
|
Ramirez, F., & Allende, H. (2013). Detection of flaws in aluminium castings: a comparative study between generative and discriminant approaches. Insight, 55(7), 366–371.
Abstract: Automatic anomaly detection has become a key issue in many engineering applications due to the increasing amount of data in need of analysis. Addressing this kind of task using pattern recognition methods requires a proper design of the learning strategy, given the reduced amount of flawed cases available for training compared to that of normal instances, which has been shown to hinder the performance of traditional classification algorithms. Moreover, positive examples are often costly and hard to collect, which may prevent the use of traditional discriminant approaches such as artificial neural networks. In this paper, we compare two well-known generative and discriminant pattern recognition algorithms in the task of flaw detection in aluminium castings and show that defects can be accurately identified without prior knowledge of positive cases, using only information of regular structures, achieving a geometric mean of over 0.9.
|
Salas, R., Allende, H., Moreno, S., & Saavedra, C. (2005). Flexible Architecture of Self Organizing Maps for changing environments. Lect. Notes Comput. Sc., 3773, 642–653.
Abstract: Catastrophic Interference is a well known problem of Artificial Neural Networks (ANN) learning algorithms where the ANN forget useful knowledge while learning from new data. Furthermore the structure of most neural models must be chosen in advance. In this paper we introduce a hybrid algorithm called Flexible Architecture of Self Organizing Maps (FASOM) that overcomes the Catastrophic Interference and preserves the topology of Clustered data in changing environments. The model consists in K receptive fields of self organizing maps. Each Receptive Field projects high-dimensional data of the input space onto a neuron position in a low-dimensional output space grid by dynamically adapting its structure to a specific region of the input space. Furthermore the FASOM model automatically finds the number of maps and prototypes needed to successfully adapt to the data. The model has the capability of both growing its structure when novel clusters appears and gradually forgets when the data volume is reduced in its receptive fields. Finally we show the capabilities of our model with experimental results using synthetic sequential data sets and real world data.
|