Swee Chiang Chiam

A competitive and cooperative co-evolutionary approach to multi-objective particle swarm optimization algorithm design

Multi-objective particle swarm optimization (MOPSO) is an optimization technique inspired by bird flocking, which has been steadily gaining attention from the research community because of its high convergence speed. On the other hand, in the face of increasing complexity and dimensionality of today’s application coupled with its tendency of premature convergence due to the high convergence speeds, there is a need to improve the efficiency and effectiveness of MOPSO. In this paper a competitive and cooperative co-evolutionary approach is adapted for multi-objective particle swarm optimization algorithm design, which appears to have considerable potential for solving complex optimization problems by explicitly modeling the co-evolution of competing and cooperating species. The competitive and cooperative co-evolution model helps to produce the reasonable problem decompositions by exploiting any correlation, interdependency between components of the problem. The proposed competitive and cooperative co-evolutionary multi-objective particle swarm optimization algorithm (CCPSO) is validated through comparisons with existing state-of-the-art multi-objective algorithms using established benchmarks and metrics. Simulation results demonstrated that CCPSO shows competitive, if not better, performance as compared to the other algorithms.

Balancing exploration and exploitation with adaptive variation for evolutionary multi-objective optimization

Although recent studies have shown that evolutionary algorithms are effective tools for solving multi-objective optimization problems, their performances are often bottlenecked by the suitability of the evolutionary operators with respect to the optimization problem at hand and their corresponding parametric settings. To adapt the search dynamic of evolutionary operation in multi-objective optimization, this paper proposes an adaptive variation operator that exploits the chromosomal structure of binary representation and synergizes the function of crossover and mutation. The overall search ability is deterministically tuned online to maintain a balance between extensive exploration and local fine-tuning at different stages of the evolutionary search. Also, the coordination between the two variation operators is achieved by means of an adaptive control that ensures an efficient exchange of information between the different chromosomal sub-structures throughout the evolutionary search. Extensive comparative studies with several representative variation operators are performed on different benchmark problems and significant algorithmic performance improvements in terms of proximity, uniformity and diversity are obtained with the incorporation of the proposed adaptive variation operator into the evolutionary multi-objective optimization process.

A memetic model of evolutionary PSO for computational finance applications

Motivated by the compensatory property of EA and PSO, where the latter can enhance solutions generated from the evolutionary operations by exploiting their individual memory and social knowledge of the swarm, this paper examines the implementation of PSO as a local optimizer for fine tuning in evolutionary search. The proposed approach is evaluated on applications from the field of computational finance, namely portfolio optimization and time series forecasting. Exploiting the structural similarity between these two problems and the non-linear fractional knapsack problem, an instance of the latter is generalized and implemented as the preliminary test platform for the proposed EA-PSO hybrid model. The experimental results demonstrate the positive effects of this memetic synergy and reveal general design guidelines for the implementation of PSO as a local optimizer. Algorithmic performance improvements are similarly evident when extending to the real-world optimization problems under the appropriate integration of PSO with EA.

Investigating technical trading strategy via an multi-objective evolutionary platform

Conventional approach in evolutionary technical trading strategies adopted the raw excess returns as the sole performance measure, without considering the associated risk involved. However, every individual has a different degree of risk averseness and thus different preferences between risk and returns. Acknowledging that these two factors are inherently conflicting in nature, this paper considers the multi-objective evolutionary optimization of technical trading strategies, which involves the development of trading rules that are able to yield high returns at minimal risk. Popular technical indicators used commonly in real-world practices are used as the building blocks for the strategies, which allow the examination of their trading characteristics and behaviors on the multi-objective evolutionary platform. While the evolved Pareto front accurately depicts the inherent tradeoff between risk and returns, the experimental results suggest that the positive correlation between the returns from the training data and test data, which is generally assumed in the single-objective approach of this optimization problem, does not necessarily hold in all cases.

Dynamic index tracking via multi-objective evolutionary algorithm

Index tracking has been gaining in popularity in recent years, as sustainable and stable yields exceeding market returns proved to be elusive. Leveraging on the search capability of evolutionary algorithm, this paper proposed a multi-objective evolutionary index tracking platform that could simultaneously optimize both tracking performance and transaction costs throughout the investment horizon and address various real-world implementation issues in index tracking. For model evaluation, a realistic instantiation of the index tracking optimization problem that accounted for stochastic capital injections, practical transactional cost structures and other real-world constraints was formulated. Portfolio rebalancing strategies for the alignment of the tracker portfolio to time-varying market conditions were investigated also. Empirical studies based on equity indices from major global markets were conducted and the results validated the tracking capability of the proposed index tracking system in out-of-sample data sets, whilst minimizing transaction costs throughout the investment horizon.

A realistic approach to evolutionary multiobjective portfolio optimization

This paper aims to address a more realistic model of the portfolio optimization problem, unlike other previous evolutionary multiobjective optimization approaches. For this purpose, an order-based representation is proposed, which can be easily extended to handle various realistic constraints like floor and ceiling constraint and cardinality constraint. Furthermore, the current experimental platform for evolutionary multiobjective portfolio optimization will be improved by introducing diversity measures and statistical analysis that are commonly used in performance assessment of multiobjective optimizers. Comparative study with other conventional representations, based on benchmark problems obtained from the OR-library, demonstrated that the proposed representation is able to attain a better approximation of the efficient frontier in terms of proximity and diversity. Experimental results also validated its viability and practicality in handling the various realistic constraints. Lastly, preference based techniques are considered also, allowing the evolutionary search to be focused on specific region of the efficient frontier. Future work includes improving the algorithmic model with more sophisticated variation operators and local search operators for better exploration and exploitation of the search space.

An Investigation on Noisy Environments in Evolutionary Multi-Objective Optimization

In addition to the need of satisfying several competing objectives, many real-world applications are also characterized by a certain degree of noise, manifesting itself in the form of signal distortion or uncertain information. While studies have shown that many multi-objective evolutionary optimizers are capable of achieving optimization goals, their ability to deal with noise is rarely studied. In this paper, extensive studies are carried out to examine the impact of noisy environments in evolutionary multi-objective optimization based upon five benchmark problems characterized by different difficulties in local optimality, non-uniformity, discontinuity and non-convexity. Interestingly, the baseline algorithm employed tends to evolve better solution sets in the presence of low noise levels for some problems. Nevertheless, the evolutionary optimization process degenerates into random search under increasing noise levels

Multiobjective evolutionary neural networks for time series forecasting

This paper will investigate the application of multiobjective evolutionary neural networks in time series forecasting. The proposed algorithmic model considers training and validation accuracy as the objectives to be optimized simultaneously, so as to balance the accuracy and generalization of the evolved neural networks. To improve the overall generalization ability for the set of solutions attained by the multiobjective evolutionary optimizer, a simple algorithm to filter possible outliers, which tend to deteriorate the overall performance, is proposed also. Performance comparison with other existing evolutionary neural networks in several time series problems demonstrates the practicality and viability of the proposed time series forecasting model.

Issues of Binary Representation in Evolutionary Algorithms

Recent studies show that evolutionary algorithms are effective optimization tools for their success in solving real-world problem with complex and competing specifications. Although their performances are greatly influenced by the type of representation adopted, this choice often arises from intuition and guesswork due to the absence of proper guidelines and framework. This paper considers binary representation and presents a comprehensive study on its issues, identifying the key factors that affect its algorithmic performance. Furthermore, two metrics are proposed to generalize the concept of preservation which quantifies the similarities between the genotype and phenotype search space. The two classical translation codes i.e. binary and gray are studied based on the identified factors and a preservation analysis revealed the differences between them

Improving Locality in Binary Representation via Redundancy

Binary representation suffers from the problem of positional dependence, where the amplitude of phenotype variation is dependent on the position of the altered genotype bits. However, this is contrary to conventional variation operations that treat each genotype bit equally. Positional dependence can be attributed to the poor locality, which results in neighboring genotypes having low correlation in the phenotype space, reducing the effectiveness of systematic local search and evolutionary search based on small mutation steps. For this purpose, this paper will propose an alternative genotype-phenotype mapping for binary representation that introduces redundancy into the mapping and removes the exponential orderings between the alleles, hence improving the locality between the genotype and phenotype search space. Empirical study conducted based on distribution, locality, and mutation innovation revealed key algorithmic characteristics of the proposed code, and its practicality is validated by comparative studies based on different benchmark optimization problems. Possible approaches to resolve the overrepresentation problem due to redundancy will be suggested, exhibiting its flexibility and variability in implementation.