Lingxi Peng

An immune-inspired semi-supervised algorithm for breast cancer diagnosis

Breast cancer is the most frequently and world widely diagnosed life-threatening cancer, which is the leading cause of cancer death among women. Early accurate diagnosis can be a big plus in treating breast cancer. Researchers have approached this problem using various data mining and machine learning techniques such as support vector machine, artificial neural network, etc. The computer immunology is also an intelligent method inspired by biological immune system, which has been successfully applied in pattern recognition, combination optimization, machine learning, etc. However, most of these diagnosis methods belong to a supervised diagnosis method. It is very expensive to obtain labeled data in biology and medicine. In this paper, we seamlessly integrate the state-of-the-art research on life science with artificial intelligence, and propose a semi-supervised learning algorithm to reduce the need for labeled data. We use two well-known benchmark breast cancer datasets in our study, which are acquired from the UCI machine learning repository. Extensive experiments are conducted and evaluated on those two datasets. Our experimental results demonstrate the effectiveness and efficiency of our proposed algorithm, which proves that our algorithm is a promising automatic diagnosis method for breast cancer.

EDA-Based multi-objective optimization using preference order ranking and multivariate gaussian copula

Estimation of distribution algorithms (EDAs) are a class of evolutionary optimization algorithms based on probability distribution model. This article extends the basic EDAs for tackling multi-objective optimization problems by incorporating multivariate Gaussian copulas for constructing probability distribution model, and using the concept of preference order. In the algorithm, the multivariate Gaussian copula is used to construct probability distribution model in EDAs. By estimating Kendalls τ and using the relationship of correlation matrix and Kendalls τ, correlation matrix R in Gaussian copula are firstly estimated from the current population, and then is used to generate offsprings. Preference order is used to identify the best individuals in order to guide the search process. The population with the current population and current offsprings population is sorted based on preference order, and the best individuals are selected to form the next population. The algorithm is tested to compare with NSGA-II, GDE, MOEP and MOPSO based on convergence metric and diversity metric using a set of benchmark functions. The experimental results show that the algorithm is effective on the benchmark functions.

Multiobjective Estimation of Distribution Algorithms Using Multivariate Archimedean Copulas and Average Ranking

Estimation of distribution algorithms (EDAs) are a class of evolutionary optimization algorithms based on probability distribution model. This chapter extends the basic EDAs for tackling multiobjective optimization problems by incorporating multivariate Archimedean copulas for constructing probability distribution model, and using the concept of average ranking. In the algorithm, Archimedean copula is used to construct probability distribution model in EDA. By estimating Kendall’s τ and using the relationship of parameter of Archimedean copula generator and Kendall’s τ, the parameter of generator in Archimedean copula are first estimated from the current population. Then Archimedean copula sampling algorithm is used to generate offsprings. Average ranking is used to identify the best individuals in order to guide search process. Population with the current population and current offsprings population is sorted based on average ranking, and the best individuals are selected to form the next population. The algorithm is tested to compare with NSGA-II, GDE, MOEP, and MOPSO based on convergence metric and diversity metric using a set of benchmark functions. Experimental results show that the algorithm is effective on the benchmark functions.

A Multi-objective PSO with Pareto Archive for Personalized E-Course Composition in Moodle Learning System

A velocity-free fully informed particle swarm optimization algorithm is firstly proposed for multi-objective optimization problems in this paper. It finds the non-dominated solutions along the search process using the concept of Pareto dominance and uses an external archive for storing them. Distinct from other multi-objective PSO, particles in swarm only have position without velocity and all personal best positions are considered to update particle position in the algorithm. The theoretical analysis implies that the algorithm will cause the swarm mean converge to the center of the Pareto optimal solution set in a multi-objective search space. Then, the algorithm is applied to the personalized e-course composition in Moodle learning system. The relative experimental results show that the algorithm has better performance and is effective.

Multi-objective cloud estimation of distribution particle swarm optimizer using maximum ranking

By combining CMBOA and PSO with quasi-oppositional learning, a cloud estimation of distribution particle swarm optimizer is firstly introduced. Then, it is extended to multi-objective optimization problems by using maximum ranking. In the algorithms offspring generation scheme, new individuals are generated in the cloud estimation of distribution way or in the PSO way. And instead of Pareto dominance, maximum ranking is used to identify the best individuals in order to guide the search process. An archive with maximum capacity is maintained. At iteration, the archive gets updated with the inclusion of the non-dominated solutions from the combined population and the archive. If the size of the archive exceeds the maximum capacity, it is cut according to the maximum ranking. The proposed algorithm is applied to some well-known benchmarks. Convergence metric, diversity metric are used to evaluate the performance of the algorithm. The experimental results show that the algorithm is effective on the benchmark functions.

Opposition-Based Learning Fully Informed Particle Swarm Optimizer without Velocity

By applying full information and employing the notion of opposition-based learning, a new opposition based learning fully information particle swarm optimiser without velocity is proposed for optimization problems. Different from the standard PSO, particles in swarm only have position without velocity and the personal best position gets updated using opposition-based learning in the algorithm. Besides, all personal best positions are considered to update particle position. The theoretical analysis for the proposed algorithm implies that the particle of the swarm tends to converge to a weighted average of all personal best position. Because of discarding the particle velocity, and using full information and opposition-based learning, the algorithm is the simpler and more effective. The proposed algorithm is applied to some well-known benchmarks. The relative experimental results show that the algorithm achieves better solutions and faster convergence.

Multi-objective opposition-based learning fully informed particle swarm optimizer with favour ranking

Some particle swarm optimization(PSO) algorithms have been proposed in recent past to tackle the multi-objective optimization problems based on the concept of Pareto optimality. In this paper, we propose a new opposition-based learning fully informed particle swarm optimizer with favour ranking to solve multi-objective optimization problems. Instead of Pareto dominance, favour ranking is used to identify the best individuals in order to guide the search process in the proposed algorithm. Different from other multi-objective PSO, particles in swarm only have position without velocity and the personal best position gets updated using opposition-based learning and favour ranking. Besides, all personal best positions are considered to update particle position. Because of discarding the particle velocity and using full information and favour ranking, the algorithm is the simpler and more effective. The proposed algorithm is applied to some well-known benchmarks. Convergence metric, diversity metric are used to evaluate the performance of the algorithm. The experimental results show that the algorithm is effective on the benchmark functions.

Archimedean copula-based estimation of distribution algorithm for multi-objective optimisation

Estimation of distribution algorithms (EDAs) are a class of evolutionary optimisation algorithms based on probability distribution model. This article extends the basic EDAs for tackling multi-objective optimisation problems by incorporating Archimedean copulas for constructing probability distribution model, and using the concept of Pareto dominance. In the algorithm, the marginal distributions from the current population are firstly estimated by kernel estimation method and are used to estimate the parameter of the Archimedean copula function generator by using the maximum likelihood method. Afterwards, the multivariate Archimedean copula sample algorithm is used to generate current offsprings population by sampling the n-dimensional Laplace transform Archimedean copula. The population with the current population and current offsprings population is sorted based on non-domination, and the best individuals are selected to form the next population based on rank and the crowding distance. The proposed algorithm is tested to compare with NSGA-II, PAES and SPEA2 using a set of benchmark functions. Both convergence and diversity metrics are used to evaluate the performance of the algorithm. The experimental results show that the algorithm outperforms NSGA-II, PAES and SPEA2 in two metrics.

Pareto-based multi-objective estimation of distribution algorithm with Gaussian copulas and application in RFID network planning

A Pareto-based multi-objective estimation of distribution algorithm with Gaussian copulas is proposed and applied to RFID network planning. The algorithm employs Pareto-based approach and multivariate Gaussian copulas to construct probability distribution model. By estimating Kendalls tau and using the relationship of Kendalls tau and correlation matrix, Gaussian copula parameters are firstly estimated, thus, joint distribution is estimated. Afterwards, the Monte Carte simulation is used to generate new individuals. An archive with maximum capacity is used to maintain the non-dominated solutions. The Pareto optimal solutions are selected from the archive on the basis of the diversity of the solutions, and the crowding-distance measure is used for the diversity measurement. The archive gets updated with the inclusion of the non-dominated solutions from the combined population and current archive, and the archive which exceeds the maximum capacity is cut using the diversity consideration. The proposed algorithm is applied to some benchmark and RFID network planning. The relative experimental results show that the algorithm has better performance and is effective.

An Immune-inspired Adaptive Automated Intrusion Response System Model

Abstract An immune-inspired adaptive automated intrusion response system model, named as MAIM, is proposed. The descriptions of self, non-self, immunocyte, memory detector, mature detector and immature detector of the network transactions, and the realtime network danger evaluation equations are given. Then, the automated response polices are adaptively performed or adjusted according to the realtime network danger. Thus, MAIM not only accurately evaluates the network attacks, but also greatly reduces the response times and response costs.