Integrated Research And Application Of DNA Genetic Algorithm

As a branch of evolutionary computation,Genetic algorithms have been extensively studied and applied in many fields,but there are still shortcomings.The local search ability of genetic algorithm is weak,and it takes a long time to solve the problem in the vicinity of the global optimal solution.After the genetic algorithm evolves to a certain extent,the individual similarity in the population can not be further of the exploration,leading to premature convergence.Not only that,the traditional genetic algorithm commonly used binary coding,which can not represent the aboundant genetic information,and can not reflect the genetic information on the growth and development of organisms,especially the key control of the role of DNA Coding mechanism.With the advent of DNA computing and development,it has been found that DNA-based intelligent systems can reflect the genetic information of organisms.This idea is conducive to the development of more powerful and can solve the more complex problems of intelligent behavior.Inspired by this idea,the researchers began to think about the possibility of further analysis and imitation of genetic information regulation system function,so as to establish the molecular level of genetic information model.Based on this idea,some researchers have proposed a DNA genetic algorithm,which is based on the DNA coding method of the individual population in the genetic operation,so as to better simulate the biological genetic mechanism and genetic information expression mechanism.The basic structure of DNA genetic algorithm is similar to genetic algorithm.The difference is that the DNA genetic algorithm uses DNA coding method to obtain the solution of the problem.Because DNA genetic algorithm is developed on the framework of genetic algorithm,it has some advantages inherent in traditional genetic algorithm,such as the operation of parameter coding,and has excellent global search performance and implicit parallelism.Compared with the traditional genetic algorithm,DNA genetic algorithm in the coding method has a greater improvement,more suitable for the expression of complex knowledge,flexible method,and high coding accuracy.As a result of the introduction of complex genetic level of operation,we can develop more and more effective genetic operators,such as inversion,separation,ectopic,etc.,thus enrich the means of evolution.In this paper,we are devoted to improve DNA genetic algorithm and integrate it with other intelligent computations in many applications,the main research work is as follows.(1)We present a new GA,called GA-TNE+DRO,which uses a novel triplet nucleotide coding scheme to encode individuals of GA and provides a set of novel genetic operators that mimic DNA molecule genetic operations.Specifically,we make the following contributions in this paper.We define a new DNA coding scheme which encodes the potential solution problem space using triplet nucleotides that represent amino acids.We define a set of evolutional operations that create new individuals in the problem space by mimicking the DNA reproduction process at an amino acid level.We perform experiments to evaluate the performance of the algorithm using a benchmark of eight unconstrained optimization problems and compare it with state-of-the-art algorithms.Our experimental results show that our algorithm can converge to solutions much closer to the global optimal solutions in a much lower number of iterations than the existing algorithms.(2)A DNA-GA algorithm based on membrane structure(m DNA-GA)is developed in this study.This method merges DNA-GA into a membrane structure.The membrane system used in this study is hierarchical composing of the skin membrane,two middle membranes and many elementary membranes.Through the collaboration of the membranes in the hierarchical membrane structure,the algorithm improves the search performance of GA.In the m DNA-GA,the elementary and middle membranes are primarily responsible for local search,and the skin membrane is for global search.The m DNA-GA algorithm starts running from the elementary membranes,and outputs the results to the middle membranes.The final,hopefully optimal,solution is obtained through the skin membrane.The m DNA-GA is applied to seven typical test optimization functions.The computational results demonstrate that m DNA-GA is very effective.(3)We construct a cloud model based DNA-GA to solve numerical optimization problems.In this study,DNA encoding scheme is adopted to encode chromosomes in genetic algorithm and cloud model is used to keep good uncertainty conversion capability and enhance function approximation ability.Based on the stochastic and stable characteristics of the normal cloud model,combined with genetic crossover and mutation,this study performs crossover operation by the Y-condition cloud generation and mutation operation by the basic cloud generator,respectively.The evolutionary process is completed skillfully.The genetic cloud operator is realized,that is,through the Y conditional cloud generator algorithm to realize crossover operator,as well as the positive normal cloud generator algorithm to update the individual,in order to achieve the evolution of the population.(4)We present a novel double-strand DNAGA for solving multi-objective optimization problems.This algorithm uses a double-strand DNA coding scheme,a set of new genetic operations,and two new rankings of non-dominated solutions to search for optimal solutions and improve solution diversity.The coding scheme and the genetic operations mimic the biological DNA structure and behavior,and achieve a better tradeoff between preserving elite individuals and diversifying the solutions.We introduce two new ranking criteria,the variant crowding distance(VCD)and the non-dominated rank with density(NRD),to help maintain diversity in the solution population.These rankings improve on the crowding distance and the non-dominated ranking and can be used to effectively identify well diversified solutions in Pareto front as well as in lateral fronts.We perform extensive experiments to compare our algorithm with several state-of-the-art MOEAs on a benchmark of wellknown bi-and tri-objective optimization problems.Preliminary results show that our algorithm outperforms the other algorithms on several performance metrics,including inverted generational distance(IGD),even spacing(ES),maximum spread(MS),convergence rate,and solution accuracy.(5)Inspired by the evolutionary clustering and DNA-GA,to overcome the drawback of k-means,we propose a DNAGA-based clustering method.In the proposed approach,the encoding is a centroid-based representation.Each individual represents a set of centroids as a k × dim dimensional vector(where k is the number of clusters and dim is the dimension of the points).The centroids of spectral graph have been employed and applied on evolving clustering,which have been transformed in such a way so as to play the role of the chromosomes in the genetic algorithm.The initial population,for the needs of the genetic algorithm,is constructed with the aid of k-nearest neighbor graphs,represented as matrices.The DNA genetic algorithm is performed based on the value of the employed fitness function that references to some of the most common clustering criteria.In the proposed method,we make use of spectral graph clustering in order to find logical grouping of the dataset.(6)We propose an automatic density peaks clustering approach using DNA genetic algorithm optimized data field and Gaussian process(referred to as ADPC-DNAGA).ADPC-DNAGA can extract the optimal value of threshold with the potential entropy of data field and automatically determine the cluster centers by Gaussian method.For any data set to be clustered,the threshold can be calculated from the data set objectively rather than the empirical estimation.The proposed clustering algorithm is benchmarked on publicly available synthetic and real-world datasets which are commonly used for testing the performance of clustering algorithms.The clustering results are compared not only with that of DPC but also with that of several well-known clustering algorithms such as Affinity Propagation,DBSCAN and Spectral Cluster.The experimental results demonstrate that our proposed clustering algorithm can find the optimal cutoff distance dc,automatically identify clusters,regardless of their shape and dimension of the embedded space,and can often outperform the comparisons.(7)We formulate an image segmentation problem as a kernel-based intuitionistic fuzzy C-means(KIFCM)clustering problem by specifying a new parametric objective function.This formulation includes a new measure for pixel local noise,a method to model fuzzy clusters as intuitionistic fuzzy sets instead of conventional fuzzy sets,and an adaptation of a kernel trick to improve performance.We propose a new DNA-based genetic algorithm to learn the KIFCM clustering.This algorithm uses a DNA coding scheme to represent individuals(i.e.,potential solutions)and a set of improved DNA genetic operator to search through the solution space for optimal solutions.Each individual encodes a set of values of the modeling parameters,including kernel parameters.While the algorithm searches for optimal set of model parameters,it also obtains the optimal IFS based fuzzy clusters.We perform empirical study by comparing our method with six existing state-of-the-art fuzzy clustering algorithms using a set of UCI data mining data sets,a set of synthetic MRI data,and a set of clinical MRI datasets.Our preliminary results show that our algorithm outperforms the compared algorithms in both the clustering metrics and computational efficiency.(8)A novel DNA encoding genetic algorithm,called KSVM-DNAGA,is proposed to search for optimal values for the parameters in kernel support vector machines.With this algorithm,the training process of support vector machines can converge quickly and the performance of the support vector ma-chines can improve.Magnetic resonance imaging(MRI)is a non-invasive diagnostic tool very frequently used for brain imaging.The classification of MRI images of normal and pathological brain conditions pose a challenge from technological and clinical point of view,since MR imaging focuses on soft tissue anatomy and generates a large information set and these can act as a mirror reflecting the conditions of the brain.The proposed method simplifies the task to a binary classification problem.We use discrete wavelet transform(DWT)to extract wavelet coefficients from MR brain images.Next,DNA-GA optimized kernel support vector machine with radial basis function(RBF)kernel,is employed as classifier.The parameters in the support vector machines are encoded into chromo-somes using DNA encoding.In the experimental procedure,four datasets are used in the computational ex-periments to verify the effectiveness of KSVM-DNAGA.We create a 90 images dataset brain downloaded from Harvard Medical School website.The 5-folded cross-validation classification results show that our method achieved 97.78% classification accuracy,higher than the comparisons.