Catálogo de publicaciones - libros

Evolutionary dynamics on graphs for the Moran process have been previously examined within the context of fixation behaviour for introduced mutants, where it was demonstrated that certain spatial structures act as amplifiers of selection. This paper will revisit the assumptions for this spatial Moran process and show that the assumption of proportional global fitness, introduced as part of the Moran process, is necessary for the amplification of selection to occur. Under the circumstances of local proportional fitness selection the amplification property no longer holds, which supports the original results from population genetics that spatial structure does not alter fixation probability for fixed population sizes with constant migration.

The performance of cooperative co-evolutionary genetic algorithms is highly affected by the representative selection strategy. But rational method is absent now. Oriented to the shortage, the representative selection strategy is studied based on the parallel implementation of cooperative co-evolutionary genetic algorithms in LAN. Firstly, the active cooperation ideology for representative selection and the dynamical determinate method on cooperation pool size are put forward. The methods for determining cooperation pool size, selecting cooperators and permuting cooperations are presented based on the evolutionary ability of sub-population and distributive performance of the individuals. Thirdly, the implementation steps are given. Lastly, the results of benchmark functions optimization show the validation of the method.

A method for object recognition of Kernel matching pursuits (KMP) [1] based on Immune Clonal algorithm (ICA) [2] is presented. Using the immune clonal select algorithm, which combines the global optimal searching ability and the locally quickly searching ability in search basic function data in function dictionary, this method can reduces computational complexity of basic matching pursuits algorithm. As compared with kernel matching pursuits the method has higher accurate recognition rate.

In this paper, a new approach is presented for the detection and classification of PQ disturbance in power system by Chirplet transforms(CT), which is the generalized forms of Fourier transform(FT), short-time Fourier transform(STFT) and wavelet transform(WT). WT and wavelet ridge are very useful tools to analyze PQ disturbance signals, but invalid for nonlinear time-varying harmonic signals. CT can detect and identify voltage quality and frequency quality visually, i.e., according to the contour of CT matrix of PQ harmonic signals, the harmonics can be detect and identify to fixed, linear time-varying and nonlinear time-varying visually. It is helpful to choose appropriate WT to analyze harmonics. Simulations show the contours of CT can effectively detect harmonic disturbance occurrence time and duration. Finally, it is validated that the harmonics of the stator current fault signal of the bar-broken electric machine is nonlinear time-varying, and tend to stable status in a short time.

The aim of this paper is twofold, to improve the generalization ability, and to improve the readability of learning classifier system. Firstly, an ensemble architecture of LCS (LCSE) is described in order to improve the generalization ability of the original LCS. Secondly, an algorithm is presented for compacting the final classifier population set in order to improve the readability of LCSE, which is an amendatory version of CRA brought by Wilson. Some test experiments are conducted based on the benchmark data sets of UCI repository. The experimental results show that LCSE has better generalization ability than single LCS, decision tree, neural network and their bagging methods. Comparing with the original population rulesets, compact rulesets have readily interpretable knowledge like decision tree, whereas decrease the prediction precision lightly.

Evolutionary Learning Classifier Systems (LCSs) are rule based systems that have been used effectively in concept learning. XCS is a prominent LCS that uses genetic algorithms and reinforcement learning techniques. In traditional machine learning (ML), early stopping has been investigated extensively to the extent that it is now a default mechanism in many systems. However, there has been a belief that EC methods are more resilient to overfitting. Therefore, this topic is under-investigated in the evolutionary computation literature and has not been investigated in LCS. In this paper, we show that it is necessary to stop evolution in LCS using a stopping criteria other than a maximum number of generations and that evolution may suffer from overfitting similar to other ML methods.

This paper introduces a computational model simulating the dynamic process of human immune response for solving Traveling Salesman Problems (TSPs). The new model is a quaternion (, , , ), where denotes exterior stimulus or antigen, denotes the set of valid antibodies, denotes the set of reaction rules describing the interactions between antibodies, and denotes the dynamic algorithm describing how the reaction rules are applied to antibody population. The set of immunodominance rules, the set of clonal selection rules, and a dynamic algorithm TSP-PAISA are designed. The immunodominance rules construct an immunodominance set based on the prior knowledge of the problem. The antibodies can gain the immunodominance from the set. The clonal selection rules strengthen these superior antibodies. The experiments indicate that TSP-PAISA is efficient in solving TSPs and outperforms a known TSP algorithm, the evolved integrated self-organizing map.

A novel strategy termed as mutation history learning strategy (MHLS) is proposed in this paper. In MHLS, a vector called mutation memory is introduced for each antibody and a new type of mutation operation based on mutation memory is also designed. The vector of mutation memory is learned from a certain antibody’s iteration history and used as guidance for its further evolution. The learning and usage of history information, which is absent from immune clonal selection algorithm (CSA), is shown to be an efficient measure to guide the direction of the evolution and accelerate algorithm’s converging speed. Experimental results show that MHLS improves the performance of CSA greatly in dealing with the function optimization problems.

This paper proposes a new immune clonal algorithm, called a quantum-inspired immune clonal algorithm (QICA), which is based on the concept and principles of quantum computing, such as a quantum bit and superposition of states. Like other evolutionary algorithms, QICA is also characterized by the representation of the antibody (individual), the evaluation function, and the population dynamics. However, in QICA, an antibody is proliferated and divided into a subpopulation. Antibodies in a subpopulation are represented by multi-state gene quantum bits. For the novel representation, we put forward the quantum mutation operator which is used at the inner subpopulation to accelerate the convergence. Finally, QICA is applied to a practical case, the multiuser detection in DS-CDMA systems, with a satisfactory result.

This paper summarises the current literature on immune system function and behaviour, including pattern recognition receptors, danger theory, central and peripheral tolerance, and memory cells. An artificial immune system framework is then presented based on the analogies of these natural system components and a rule and feature-based problem representation. A data set for intrusion detection is used to highlight the principles of the framework.