Application of imperialist competitive algorithm to find minimax and standardized maximin optimal designs
Introduction
A wide class of evolutionary algorithms has been increasingly used to solve hard optimization problems in engineering, bioinformatics, computer science and finance. Of particular interest is the class of nature-inspired algorithms motivated from the influence of biology and the life sciences. Examples of such algorithms frequently used in the statistics literature are simulated annealing (SA) and genetic algorithms (GA). There are more recent and potentially more powerful ones such as differential evolution (DE), particle swarm optimization (PSO), imperialist competitive algorithm (ICA) and cuckoo search (CS) that have not yet been well tested for solving statistical problems. Our interest in this paper is application of one such algorithm to construct optimal experimental designs. The usefulness of implementing such designs in terms of cost saving and accurate statistical inferences is detailed in Atkinson (1996). An introduction to this subfield of optimal experimental designs is available in Berger and Wong (2009) and real applications of optimal designs can be found in Berger and Wong (2005).
A common appeal of such algorithms is that they are mainly assumptions free, fast, easy to implement and are broadly applicable to different types of constrained or unconstrained optimization problems. Consequently, they have good potential to optimize complicated functions with many variables regardless whether the objective function is differentiable or not. A common feature among these algorithms is that there require tuning parameters and if they are well chosen, the algorithm finds the optimum very fast. If the tuning parameters are poorly chosen, the algorithm does not give satisfactory answers. These algorithms do not usually have a firm theoretical basis, such as proof of its convergence to the optimum. However, these algorithms have been used successfully in many applied fields to solve real, complicated and high dimensional optimization problems where traditional formulations or methods fail. Our view is that the lack of proof of convergence should not hinder their use in statistics; for some problems such as the ones we work with, there is a theory for verifying whether the generated design is optimum, and if it is not, theory is available to assess its proximity to the optimum without knowing the optimum. We next briefly review selected applications of such algorithms for finding optimal designs in the literature.
Simulated annealing was first proposed by Kirkpatrick et al. (1983) and Meyer and Nachtsheim (1988) appeared to be among the first to use SA and constructed exact -optimal designs for both finite and continuous design spaces. Haines (1987) applied SA to construct exact -, -, and -optimal designs for polynomial regression models. Atkinson (1992) discussed starting values for SA with a focus on optimal design construction and recommended segmenting the search to a maximum number of evaluations. Other applications of SA include Zhou (2008), who found exact minimax -optimal design on discrete design spaces, and Wilmut and Zhou (2011), who constructed -optimal minimax two-level fractional factorial designs using a sequential algorithm. Woods (2010) implemented SA to obtain exact optimal designs for binary models under the optimum-in-average criterion. This algorithm is one of the very few that can be shown to converge to the optimum.
GA was proposed by Holland and John (0000) and has been applied to search for exact optimal designs mainly for linear models. An early proponent is Montepiedra et al. (1998), who found exact optimal designs for polynomial models; others include Heredia-Langner et al. (2003), Drain et al. (2004) and Mandal et al. (2015). Hamada et al. (2001) used GA to find near-optimal Bayesian experimental designs for linear and nonlinear regression and dichotomous data. A most recent review on the application of GA to solve optimal design problems is given in Lin et al. (2015).
Particle swarm optimization (PSO) algorithm proposed by Eberhart and Kennedy (1995) has emerged to be a popular tool for solving real world optimization problems. In PSO terminology, each possible solution is called a “particle”. Similar to many evolutionary algorithms, PSO is initialized with a population of random particles, called a “swarm”. Each particle starts to fly through the problem space with its own “velocity” that is being updated in every iteration according to the particle’s best position and the best global position over the swarm. PSO has been used to find several types of optimal designs for different problems. For example, Qiu et al. (2014) applied PSO to find locally - and -optimal designs for the compartmental, logistic and double exponential models and comparing PSO performance with the differential evolution algorithm proposed by Storn and Price (1997). Wong et al. (2015) demonstrated the usefulness of using PSO to find various types of optimal designs by applying it to solve several types of optimal design problems for different mixture models defined over a regular or irregular simplex. In addition, Chen et al. (2014) modified PSO to find minimax optimal designs for the logistic and enzyme kinetic models. Such optimal designs are notoriously difficult to find because the design criterion is non-differentiable and involves two layers of optimization.
The main goal of this paper is to investigate the capability of the Imperialist Competitive Algorithm (ICA) for finding optimal designs. Our work appears to be the first to use ICA for a statistical application. ICA is a meta-heuristic evolutionary algorithm inspired from the socio-political process of humans and proposed in Atashpaz-Gargari and Lucas (2007). In this sense, it is different from the above mentioned nature-inspired algorithms, which are inspired by animal behavior. ICA has been successfully applied in engineering subfields such as industrial, civil, mechanical, electronic, petroleum and computer engineering; see Hosseini and Al Khaled (2014) for a review. Our interest in ICA is in part due to recent reports from the engineering literature that suggest ICA can outperform some widely used evolutionary nature-inspired algorithms including PSO. For example, Hamel et al. (2012) compared performances of ICA with PSO to optimize some famous test functions like the De Jong’s, Rastrigin’s and Hartmann’s functions in nondestructive Eddy-Current Testing (ECT) problems. They reported that when the objective function has five or fewer parameters, ICA and PSO techniques performed almost the same. However, when the number of parameters was increased, ICA found the solution faster than PSO and with more accuracy (see Table 3 of Hamel et al., 2012 for details).
In what is to follow, we focus on finding minimax type of optimal designs using ICA. These are hard design problem because they involve solving nested multi-level optimization problems, the optimality criterion is not differentiable and there is no algorithm that we know of that can be shown to converge to such optimal designs in a general nonlinear regression setup. We show how to modify ICA with a perturbed move, coupled with a local search procedure to find the optimal designs more effectively.
In the next section, we review the statistical setup and theory for finding optimal designs. We present details and implementation information for the ICA in Section 3. In Section 4 we demonstrate how ICA may be applied to find minimax optimal designs for the power logistic model. Section 5 modifies ICA to find standardized maximin -optimal designs for the log-linear, exponential and enzyme kinetic models. Section 6 provides a discussion.
Section snippets
Background and minimax optimal designs
Throughout we focus on approximate designs proposed by Kiefer in late 1950s. His subsequent work and numerous applications of approximate designs ideas are now voluminously documented in Kiefer (1985). An approximate design is a probability measure defined on a user-selected design space . Let be the space of all such designs on and let be an approximate design with support points at from with corresponding weights , , . This means that when we have a
The Imperialist Competitive Algorithm (ICA)
This section describes the basic concepts behind ICA and we show they may be used to search for optimal designs. ICA begins with a random population of solutions called countries. Countries within ICA are analogous to particles in PSO. The population of countries is divided into some sub-populations called empires. Each empire contains one imperialist and some colonies. These empires start to evolve simultaneously by moving all colonies toward the imperialists through an operation called
Minimax -optimal designs for the power logistic model
The power logistic model is given by where . In the minimax approach, we suppose belongs to the region of uncertainty , where are known limits of the lower and upper bounds for and . This model is sometimes called the 3-parameter logistic model and is more flexible than the simple logistic model where . When , heteroscedasticity and skewness can be accommodated by the power logistic model but not the simple logistic
Standardized maximin -optimal designs
Minimax optimal designs are essentially equivalent to maximin optimal designs. These optimal designs can have very different criterion values depending on the nominal set of parameter values. Accordingly, it is desirable to standardize the criterion and control for the potentially widely varying magnitude of the criterion.
A design is a standardized maximin -optimal design (Dette, 1997) if where is the locally -optimal design for the parameter
Discussion
In this paper, we proposed and tested ICA as an alternative evolutionary algorithm to find minimax and standardized maximin optimal designs for nonlinear models. We applied the ICA algorithm and generated minimax optimal designs for the power logistic model, standardized maximin -optimal designs for the log-linear model, exponential model and several inhibition models useful in enzyme kinetic studies. These are hard problems because the design criterion is non-differentiable and requires
Acknowledgments
The authors would like to thank three anonymous referees and an Associate Editor for helpful comments and suggestions on an earlier version of the article. The research of Wong reported in this paper was partially supported by a grant award R01GM107639 from the National Institute of General Medical Sciences of the National Institutes of Health. The contents in this paper are solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes
References (62)
- et al.
Optimal Bayesian design applied to logistic regression experiments
J. Statist. Plann. Inference
(1989) - et al.
Bayesian D-optimal designs for exponential regression models
J. Statist. Plann. Inference
(1997) - et al.
E-optimal designs for the Michaelis–Menten model
Statist. Probab. Lett.
(1999) - et al.
A survey on the imperialist competitive algorithm metaheuristic: implementation in engineering domain and directions for future research
Appl. Soft Comput.
(2014) - et al.
Optimum design of skeletal structures using imperialist competitive algorithm
Comput. Struct.
(2010) - et al.
Using animal instincts to design efficient biomedical studies via particle swarm optimization
Swarm Evol. Comput.
(2014) - et al.
Imperialist competitive algorithm combined with chaos for global optimization
Commun. Nonlinear Sci. Numer. Simul.
(2012) - et al.
D-optimal minimax design criterion for two-level fractional factorial designs
J. Statist. Plann. Inference
(2011) D-optimal minimax regression designs on discrete design space
J. Statist. Plann. Inference
(2008)- et al.
Experimental design for binary data
J. Amer. Statist. Assoc.
(1983)
Imperialist competitive algorithm: an algorithm for optimization inspired by imperialistic competition
The usefulness of optimum experimental designs
J. R. Stat. Soc. Ser. B Stat. Methodol.
A segmented algorithm for simulated annealing
Stat. Comput.
Optimum Experimental Designs, with SAS, Vol. 34
Optimal and efficient designs of experiments
Ann. Math. Statist.
Minimax D-optimal designs for item response theory models
Psychometrika
An Introduction to Optimal Designs for Social and Biomedical Research
Applied Optimal Designs
Optimum design of experiments for enzyme inhibition kinetic models
J. Biopharm. Statist.
Practical considerations for optimal designs in clinical dose finding studies
Stat. Med.
Minimax optimal designs via particle swarm optimization methods
Stat. Comput.
Locally optimal designs for estimating parameters
Ann. Math. Statist.
Designing experiments with respect to standardized optimality criteria
J. R. Stat. Soc. Ser. B Stat. Methodol.
Robust and efficient designs for the Michaelis–Menten model
J. Amer. Statist. Assoc.
Optimal designs for the EMAX, log-linear and exponential models
Biometrika
Locally D-optimal designs for exponential regression models
Statist. Sinica
A genetic algorithm hybrid for constructing optimal response surface designs
Qual. Reliab. Eng. Int.
A semi-infinite programming based algorithm for finding minimax optimal designs for nonlinear models
Stat. Comput.
Cited by (20)
Numerical Methods for Finding A-optimal Designs Analytically
2023, Econometrics and StatisticsCitation Excerpt :We mention two other nature-inspired metaheuristic algorithms that were recently used to search for optimal designs. They are the imperial competitive algorithm (ICA) (Masoudi et al., 2017; 2019) and differential evolution (DE) (Xu et al., 2019; Stokes et al., 2020). ICA is based on human behavior and DE has similar characteristics as Genetic Algorithm, which is likely more well known among statisticians.
An algorithm based on semidefinite programming for finding minimax optimal designs
2018, Computational Statistics and Data AnalysisCitation Excerpt :Qiu (2014) and Chen et al. (2015) applied a well known member of this class of algorithms, called Particle Swarm Optimization (PSO) to generate several types of optimal designs. Masoudi et al. (2017) adopted another metaheuristic procedure called Imperialist Competitive Algorithm (ICA) for the same purpose. We did not compare performance of such algorithms with our proposed approach because we feel that comparing performance of algorithms should always be done meaningfully and fairly and we do not feel it is feasible to do so here.
Sequential Design of Experiments to Maximize Learning from Carbon Capture Pilot Plant Testing
2018, Computer Aided Chemical EngineeringCitation Excerpt :It incorporates the results from the first stage to determine the location of the next experiment. For the space-filling design, the minimax criterion (Masoudi et al., 2017) was used. This objective seeks to minimize the maximum distance between any location in the design space of interest and where data have already been obtained.
Special issue on Design of Experiments
2017, Computational Statistics and Data Analysis