Why parameter control mechanisms should be benchmarked against random variation

Parameter control mechanisms in evolutionary algorithms (EAs) dynamically change the values of the EA parameters during a run. Research over the last two decades has delivered ample examples where an EA using a parameter control mechanism outperforms its static version with fixed parameter values. However, very few have investigated why such parameter control approaches perform better. In principle, it could be the case that using different parameter values alone is already sufficient and EA performance can be improved without sophisticated control strategies. This paper investigates whether very simple random variation in parameter values during an evolutionary run can already provide improvements over static values.

Prediction of stand attributes of even-aged teak stands using multilevel models

Differential evolution (DE) is one of the most successful evolutionary algorithms (EAs) for global numerical optimization. Like other EAs, maintaining population diversity is important for DE to escape from local optima and locate a near-global optimum. Using a multi-population algorithm is a representative method to avoid early loss of population diversity. In this paper, we propose a multi-population DE algorithm (MPDE) which manipulates multiple sub-populations. Different sub-populations in MPDE exchange information via a novel mutation operation instead of migration used in most multi-population EAs. The mutation operation is helpful to balance the fast convergence and population diversity of the proposed algorithm. Moreover, the performance of MPDE is further improved by an adaptive parameter control scheme designed based on the multi-population approach. Each sub-population in MPDE evolves with its own set of control parameters, and a learning strategy is used to adaptively adjust the parameter values. A set of benchmark functions is used to test the proposed MPDE algorithm. The experimental results show that MPDE performs better than, or at least comparably, to the classical single population DE with fixed parameter values and three existing state-of-the-art DE variants.

It is common to distinguish three main senses of the term “robustness”: (1) Robustness of models; (2) Robustness as stability or insensitivity of output as against variations in parameter values; (3) Robustness as consilience of results from different and independent hypotheses, procedures or sources of evidence. The purpose of this paper is to discuss the last two meanings of robustness, in order to cope with some difficulties with which robustness as consilience is confronted and which have indirect consequences for the problem of scientific realism. On the one hand, robustness regarded as reproducible stability as against perturbations and variations in parameter values (robustness-as-stability) and robustness as consilience of results from different and independent pieces of evidence (robustness-as-consilience) are conceptually distinct. On the other hand, however, robustness-as-stability is a condition of robustness-as-consilience; and the converse holds also: robustness-as-consilience is an essential ingredient of robustness-as-stability. There is no vicious circle here, but a technical-practical synergy, which is at the heart of the experimental method, and which can help us out of the two main problems for robustness-as-consilience.

In this paper, we propose a population-based, four-step, real-parameter optimization algorithm-generator. The approach divides the task of reaching near the optimum solution into four independent plans of (i) selecting good solutions from a solution base, (ii) generating new solutions using the selected solutions, (iii) choosing inferior or spurious solutions for replacement, and (iv) updating the solution base with good new or old solutions. Interestingly, many classical and evolutionary optimization algorithms are found to be representable by this algorithm-generator. The paper also recommends an efficient optimization algorithm with the possibility of using a number of different recombination plans and parameter values. With a systematic parametric study, the paper finally recommends a real-parameter optimization algorithm which outperforms a number of existing classical and evolutionary algorithms. To extend this study, the proposed algorithm-generator can be utilized to develop new and more efficient population-based optimization algorithms. The treatment of population-based classical and evolutionary optimization algorithms identically through the proposed algorithm-generator is the main hall-mark of this paper and should enable researchers from both classical and evolutionary fields to understand each other’s methods better and interact in a more coherent manner.

The wireless body area network (WBAN) has gaining tremendous attention among researchers and academicians for its envisioned applications in healthcare service. Ultra wideband (UWB) radio technology is considered as excellent air interface for communication among body area network devices. Characterisation and modelling of channel parameters are utmost prerequisite for the development of reliable communication system. The path loss of on-body UWB channel for each frequency band defined in IEEE 802.15.6 standard is experimentally determined. The parameters of path loss model are statistically determined by analysing measurement data. Both the line-of-sight and non-line-of-sight channel conditions are considered in the measurement. Variations of parameter values with the size of human body are analysed along with the variation of parameter values with the surrounding environments. It is observed that the parameters of the path loss model vary with the frequency band as well as with the body size and surrounding environment. The derived parameter values are specific to the particular frequency bands of IEEE 802.15.6 standard, which will be useful for the development of efficient UWB WBAN system.

A possible goal in robust design of dynamic systems is to find a system topology under which the sensitivity of performance to the values of component parameters is minimized. This can provide robust performance in the face of environmental change (e.g., resistance variation with temperature) and/or manufacturing-induced variability in parameter values. In some cases, a topology that is relatively insensitive to parameter variation may allow use of less expensive (looser tolerance) components. Cost of components, in some instances, also depends on whether “standard-sized” components may be used or custom values are required. This is true whether the components are electrical components, mechanical fasteners, or hydraulic fittings. However, using only standard-sized or preferred-value components introduces an additional design constraint. This chapter uses genetic programming to develop bond graphs specifying component topology and parameter values for an example task, designing a passive analog low-pass filter with fifth-order Bessel characteristics. It explores three alternative design approaches. The first uses “standard” GP and evolves designs in which components can take on arbitrary values (i.e., custom design). The second approach adds random noise to each parameter; then, at the end of evolution, for the best design found, it “snaps” its parameter values to a small (component-specific) set of “standard” values. The third approach uses only the small set of allowable standard values throughout the evolutionary process, evaluating each design after addition of noise to each standard parameter value. Then the best designs emerging from each of these three procedures are compared for robustness to parameter variation, evaluating each of them with random perturbations of their parameters. Results indicated that, the third method produced the most robust designs, and the second method was better than the first.

The aim of this presentation is to construct a robust optimal PID feedback controller, taking into account stochastic uncertainties in the initial conditions. Usually, a precomputed feedback control is based on exactly known model parameters. However, in practice, often exact information about model parameters and initial values is not given. Hence, having an inital point, which differs from the nominal values, a standard precomputed controller may produce bad results. Supposing now that the probability distribution of the random parameter variations is known, in the following stochastic optimisation methods will be applied in order to obtain robust optimal feedback controls. Taking into account stochastic parameter variations at the initial point, the method works with expected total costs arising from the primary control expenses and the tracking error. Furthermore, the free regulator parameters are selected then such that the expected total costs are minimized. After Taylor expansion to calculate expected cost functions and a few transformations an approximate deterministic substitute control problem follows. Here, retaining only linear terms, approximation of expectations and variances of the expected cost functions can be calculated explicitly. By means of splines, numerical approximations of the objective function and the differential equations are obtained then. Using stochastic optimization methods, random parameter variations are incorporated into the optimal control process. Hence, robust optimal feedback controls are obtained. (© 2009 Wiley‐VCH Verlag GmbH & Co. KGaA, Weinheim)

Abstract-The effects of random variations of the plant parameters in multivariable linear control systems are analyzed in terms of stochastic sensitivity operators. Both tie-varying and time-invariant control systems are analyzed. Conditions are derived for the stochastic sensitivity operator which are sufficient to guarantee that the closed-loop realization of the control system is less sensitive to the random parameter variations than the nominally equivalent open-loop system.

Evolutionary algorithms (EA) are optimization algorithms inspired by the neo-dar winian theory of evolution. Those algorithms use a population of potential solutions which is gradually guided towards better solutions discovered by random variation. The neutral theory of evolution considers that the majority of mutations are selectively neutral or lethal. In this article we present the main results about neutral theory in the context of EA, and particulary how this theory affects studies of fitness landscapes. Fitness landscapes represent the set of potential solutions of a problem as an space equipped with a neighborhood relation and where the solutions have an height corresponding to their performance. This article will show academic and real problems related to neutral landscapes and the nature of EA dynamics on those landscapes. In order to improve the performances of EA, we then present technics based on the choice of redundant coding, just as the design of EA exploiting neutrality.

We show how you can carry out systematic sensitivity analysis (SSA) with respect to parameters and/or shocks, which vary according to a specified covariance matrix. You can use the existing SSA tools in RunGTAP or RunGEM to do this if your model is implemented in GEMPACK. Those SSA tools assume that all parameters or shocks are varying independently (i.e., the distributions of all parameters or shocks are uncorrelated) or together (i.e., are completely correlated). The techniques in this paper remove those restrictions. However, users need to make small modifications to the TAB file for the model. Different modifications are needed for different SSA scenarios. Further, the standard SSA procedure built into RunGTAP and RunGEM allows you to compute the sensitivity of model results either with respect to variations in parameter values or with respect to variations in shock values, but you cannot vary both parameters and shocks at the same time. Our discussion concentrates on the parameter case. However, we later show how shock variation may be modelled as a type of parameter variation. This opens the door to simultaneous variation of shocks and parameters. We include worked examples of the techniques described, based on the standard GTAP Model.

The contribution is focused on the random variation and correlation of input parameters for the climate data description. The climate data such as ambient temperature, solar intensity, wind speed and direction are of the random nature. The description of ambient temperature can be based on the climate data time series in the form of climatic “load duration curve”. Particular input parameters such as ambient temperature and solar radiation have a significant correlation. The attention is paid especially to the evaluation and simulation application of the histograms using Monte Carlo type process considering correlation of particular parameters.

All types of passive components present variations of most parameters as function of the operating temperature. Depending on the component type and thermal management, in a specific application, these parameter variations can sometimes exceed 20% over the entire temperature range, having an important effect on the overall performance of the circuit. These variations in parameter values, although expected, are rarely implemented in the electric simulations because of the lack of temperature dependent electrical models for passive components. This paper presents a method and its software implementation for generating improved SPICE models for passive components, models that describe thermal variations of the main parameters based on manufacturer datasheet information, general package parasitic elements and thermal behavior measurements.

In this article, a computationally-efficient approach is presented for Bit Error Rate (BER) calculations in multi-user Multiple Input Multiple Output (MIMO) frequency selective fading environments, operating in spatial multiplexing transmission mode. To this end, theoretical expressions for the equivalent Signal to Interference plus Noise Ratio (SINR) per active user and transmission mode, as well as mean BER, are derived. The key idea is that all parameters related to BER calculations can be expressed as sums of identically distributed random variables (RVs). Hence, computational burden can be reduced, since summation formulas take into account the parameters of a standalone RV along with the correlation of an arbitrary pair of RVs. As results indicate, the proposed approach can accurately estimate mean BER in multiuser MIMO orientations with increased reception diversity order, for arbitrary number of transmit/receive antennas, a modulation scheme, and a number of resolvable multi-path components.

Polyurethane is used for making mould in soft tooling (ST) process for producing wax/plastic components. These wax components are later used as pattern in investment casting process. Due to low thermal conductivity of polyurethane, cooling time in ST process is long. To reduce the cooling time, thermal conductive fillers are incorporated into polyurethane to make composite mould material. However, addition of fillers affects various properties of the ST process, such as stiffness of the mould box, rendering flow-ability of melt mould material, etc. In the present work, multi-objective optimization of various conflicting objectives (namely maximization of equivalent thermal conductivity, minimization of effective modulus of elasticity, and minimization of equivalent viscosity) of composite material are conducted using evolutionary algorithms (EAs) in order to design particle-reinforced polyurethane composites by finding the optimal values of design parameters. The design parameters include volume fraction of filler content, size and shape factor of filler particle, etc. The Pareto-optimal front is targeted by solving the corresponding multi-objective problem using the NSGA-II procedure. Then, suitable multi-criterion decision-making techniques are employed to select one or a small set of the optimal solution(s) of design parameter(s) based on the higher level information of the ST process for industrial applications. Finally, the experimental study with a typical real industrial application demonstrates that the obtained optimal design parameters significantly reduce the cooling time in soft tooling process keeping other processing advantages.