Jesse B. Zydallis

Considerations in engineering parallel multiobjective evolutionary algorithms

Developing multiobjective evolutionary algorithms (MOEAs) involves thoroughly addressing the issues of efficiency and effectiveness. Once convinced of an MOEAs effectiveness the researcher often desires to reduce execution time and/or resource expenditure, which naturally leads to considering the MOEAs parallelization. Parallel MOEAs (pMOEAs) or distributed MOEAs are relatively new developments with few associated publications. pMOEA creation is not a simple task, involving analyzing various parallel paradigms and associated parameters. Thus, a thorough discussion of the major parallelized MOEA paradigms is included in this paper and succinct observations are made regarding an analysis of the current literature. Specifically, a previous MOEA notation is extended into the pMOEA domain to enable precise description and identification of various sets of interest. Innovative concepts for pMOEA migration, replacement and niching schemes are discussed, as well as presenting the first known generic pMOEA formulation. Taken together, this papers analyses in conjunction with an original pMOEA design serve as a pedagogical framework and example of the necessary process to implement an efficient and effective pMOEA.

Issues in parallelizing multiobjective evolutionary algorithms for real world applications

The concepts of efficiency and effectiveness must be addressed in conducting research into using a Evolutionary Algorithm (EA) for optimization problems. The increased use of evolutionary approaches for real-world applications, containing multiple objectives and high dimensionality, has led to the design and generation of a number of Multiobjective Evolutionary Algorithms (MOEA). When analyzing these algorithms, the issues of effectiveness and efficiency are extremely important and typically drive the urge to parallelize these algorithms. The parallelization of MOEAs is a relatively new concept, with few researchers contributing work in this area. This parallelization process is not a simple task and involves the analysis of various parallel models and the parameters associated with these models. This paper presents a thorough analysis of the various parallel MOEA models, the issues associated with these models and recommendations for using these models in MOEAs. In particular, these parallelization concepts are applied to the Multiobjective Messy Genetic Algorithm II.

Explicit building-block multiobjective evolutionary algorithms for NPC problems

This research emphasizes explicit building block (BB) based MOEA performance with detailed symbolic representations. An explicit BB-based MOEA for solving constrained and real-world multiple objective problems (MOPs) is developed, the multiobjective messy genetic algorithm II (MOMGA-II) in order to validate symbolic BB concepts. This algorithm provides insight into solving difficult NP-complete MOPs that are generally not realized through the use of implicit BB-based MOEA approaches. Specific constrained integer problem examples include advanced logistics and modified knapsack problems. A primary focus is on generic repair mechanisms for generating feasible solutions per generation. The insight provided is necessary to increase the effectiveness and efficiency over all possible MOEA approaches.

Solving of discrete multiobjective problems using an evolutionary algorithm with a repair mechanism

The solving of real-world multiobjective problems (MOPs) with an evolutionary algorithm (EA) is an increasing area of interest Presented in this paper is the application of a building block based EA to a real-world discrete MOP. A constraint handling method had to be designed and employed. The description of this method and the repair mechanism instrumental in repairing infeasible solutions is described in detail along with statistical analysis.

Use of mendelian pressure in a multi-objective genetic algorithm

Significant work has been conducted in developing techniques for multi-objective problem (MOP) optimizations. This paper investigates the use of a Mendel-like dominance scheme for improving the efficiency of a MOP genetic algorithm. This paper shows, for the selected MOP test suite, that the mendelian GA outperforms a simple GA.

Analysis of fine granularity and building block sizes in the parallel fast messy GA

This paper presents two methods designed to improve the efficiency and effectiveness of the parallel fast messy GA used in solving the Protein Structure Prediction (PSP) problem. The first is an application of a farming model - targeting algorithm efficiency. The second successful method addresses the building block sizes used in the algorithm - targeting algorithm effectiveness.