Feature Selection to Improve Generalization of Genetic Programming for High-Dimensional Symbolic Regression

When learning from high-dimensional data for symbolic regression (SR), genetic programming (GP) typically could not generalize well. Feature selection, as a data preprocessing method, can potentially contribute not only to improving the efficiency of learning algorithms but also to enhancing the generalization ability. However, in GP for high-dimensional SR, feature selection before learning is seldom considered. In this paper, we propose a new feature selection method based on permutation to select features for high-dimensional SR using GP. A set of experiments has been conducted to investigate the performance of the proposed method on the generalization of GP for high-dimensional SR. The regression results confirm the superior performance of the proposed method over the other examined feature selection methods. Further analysis indicates that the models evolved by the proposed method are more likely to contain only the truly relevant features and have better interpretability.

Comments Off on Feature Selection to Improve Generalization of Genetic Programming for High-Dimensional Symbolic Regression

DMOEA- $varepsilon text{C}$ : Decomposition-Based Multiobjective Evolutionary Algorithm With the $varepsilon $ -Constraint Framework

Decomposition is an efficient and prevailing strategy for solving multiobjective optimization problems (MOPs). Its success has been witnessed by the multiobjective evolutionary algorithm MOEA/D and its variants. In decomposition-based methods, an MOP is decomposed into a number of scalar subproblems by using various scalarizing functions. Most decomposition schemes adopt the weighting method to construct scalarizing functions. In this paper, another classical generation method in the field of mathematical programming, that is the $ {varepsilon }$ -constraint method, is adopted for the multiobjective optimization. It selects one of the objectives as the main objective and converts other objectives into constraints. We incorporate the $ {varepsilon }$ -constraint method into the decomposition strategy and propose a new decomposition-based multiobjective evolutionary algorithm with the $ {varepsilon }$ -constraint framework (DMOEA- $ {varepsilon }text{C}$ ). It decomposes an MOP into a series of scalar constrained optimization subproblems by assigning each subproblem with an upper bound vector. These subproblems are optimized simultaneously by using information from neighboring subproblems. Besides, a main objective alternation strategy, a solution-to-subproblem matching procedure, and a subproblem-to-solution matching procedure are proposed to strike a balance between convergence and diversity. DMOEA- $ {varepsilon }text{C}$ is compared with a number of state-of-the-art multiobjective evolutionary algorithms. Experimental studies demonstrate that DMOEA- $ {-
arepsilon }text{C}$
outperforms or performs competitively against these algorithms on the majority of 34 continuous benchmark problems, and it also shows obvious advantages in solving multiobjective 0-1 knapsack problems.

Comments Off on DMOEA- $varepsilon text{C}$ : Decomposition-Based Multiobjective Evolutionary Algorithm With the $varepsilon $ -Constraint Framework

IEEE Transactions on Evolutionary Computation Society Information

Provides a listing of current committee members and society officers.

Comments Off on IEEE Transactions on Evolutionary Computation Society Information

A Weighted Biobjective Transformation Technique for Locating Multiple Optimal Solutions of Nonlinear Equation Systems

Due to the fact that a nonlinear equation system (NES) may contain multiple optimal solutions, solving NESs is one of the most important challenges in numerical computation. When applying evolutionary algorithms to solve NESs, two issues should be considered: 1) how to transform an NES into a kind of optimization problem and 2) how to develop an optimization algorithm to solve the transformed optimization problem. In this paper, we tackle the first issue by transforming an NES into a weighted biobjective optimization problem. By the above transformation, not only do all the optimal solutions of an original NES become the Pareto optimal solutions of the transformed biobjective optimization problem, but also their images are different points on a linear Pareto front in the objective space. In addition, we suggest an adaptive multiobjective differential evolution, the goal of which is to effectively locate the Pareto optimal solutions of the transformed biobjective optimization problem. Once these solutions are found, the optimal solutions of the original NES can also be obtained correspondingly. By combining the weighted biobjective transformation technique with the adaptive multiobjective differential evolution, we propose a generic framework for the simultaneous locating of multiple optimal solutions of NESs. Comprehensive experiments on 38 NESs with various features have demonstrated that our framework provides very competitive overall performance compared with several state-of-the-art methods.

Comments Off on A Weighted Biobjective Transformation Technique for Locating Multiple Optimal Solutions of Nonlinear Equation Systems

IEEE World Congress on Computational Intelligence

Describes the above-named upcoming conference event. May include topics to be covered or calls for papers.

Comments Off on IEEE World Congress on Computational Intelligence

A Surrogate Assisted Approach for Single-Objective Bilevel Optimization

Bilevel optimization refers to a hierarchical problem in which optimization needs to be performed at two nested levels, namely the upper level and the lower level. The aim is to identify the optimum of the upper level problem, subject to optimality of the corresponding lower level problem. Several problems from the domain of engineering, logistics, economics, and transportation have inherent nested structure which requires them to be modeled as bilevel optimization problems. Bilevel optimization usually requires inordinate amount of function evaluations since a lower level search needs to be conducted for evaluating each upper level solution. The evaluations are especially high when the problems are not suited for exact techniques and evolutionary techniques are employed instead. Reducing this computational effort has been one of the key pursuits in this domain recently. However, the use of surrogate modeling to achieve this goal has so far been scarcely studied. In this paper, we present a surrogate assisted optimization approach toward addressing this research gap. The approach uses surrogates of multiple types in order to provide flexibility of approximating different types of functions more accurately. The algorithm is further strengthened through the use of selective re-evaluation of promising solutions and periodic nested local search. The performance of the proposed algorithm is presented on twenty five standard benchmark problems. The results are compared with a number of other established evolutionary and hybrid algorithms to demonstrate the efficacy of the proposed approach in obtaining competitive results using relatively fewer function evaluations.

Comments Off on A Surrogate Assisted Approach for Single-Objective Bilevel Optimization

Table of contents

Presents the table of contents for this issue of the publication.

Comments Off on Table of contents

A Two-Phase Differential Evolution for Uniform Designs in Constrained Experimental Domains

In many real-world engineering applications, a uniform design needs to be conducted in a constrained experimental domain that includes linear/nonlinear and inequality/equality constraints. In general, these constraints make the constrained experimental domain small and irregular in the decision space. Therefore, it is difficult for current methods to produce a predefined number of samples and make the samples distribute uniformly in the constrained experimental domain. This paper presents a two-phase differential evolution for uniform designs in constrained experimental domains. In the first phase, considering the constraint violation as the fitness function, a clustering DE is proposed to guide the population toward the constrained experimental domain from different directions promptly. As a result, a predefined number of samples can be obtained in the constrained experimental domain. In the second phase, maximizing the minimum Euclidean distance among samples is treated as another fitness function. By optimizing this fitness function, the samples produced in the first phase can be scattered uniformly in the constrained experimental domain. The performance of the proposed method has been tested and compared with another state-of-the-art method. Experimental results suggest that our method is significantly better than the compared method in the uniform designs of a new type of automotive crash box and five benchmark test problems. Moreover, the proposed method could be considered as a general and promising framework for other uniform designs in constrained experimental domains.

Comments Off on A Two-Phase Differential Evolution for Uniform Designs in Constrained Experimental Domains

Improving Evolutionary Algorithms in a Continuous Domain by Monitoring the Population Midpoint

It is advocated that monitoring the population midpoint allows for improving the efficiency of population-based evolutionary algorithms (EAs) in $mathbb {R}^{ d}$ . The theoretical motivation supporting this hypothesis is provided in this letter, and this phenomenon is empirically confirmed for selected typical EAs by a series of tests for fitness functions contained in the CEC2005 and CEC2013 benchmark sets.

Comments Off on Improving Evolutionary Algorithms in a Continuous Domain by Monitoring the Population Midpoint

IEEE Transactions on Evolutionary Computation publication information

Presents a listing of the editorial board, board of governors, current staff, committee members, and/or society editors for this issue of the publication.

Comments Off on IEEE Transactions on Evolutionary Computation publication information