Parameter setting and statistical learning

Three main models of parameter setting have been proposed: the Variational model proposed by Yang (2002; 2004), the Structured Acquisition model endorsed by Baker (2001; 2005), and the Very Early Parameter Setting (VEPS) model advanced by Wexler (1998). The VEPS model contends that parameters are set early. The Variational model supposes that children employ statistical learning mechanisms to decide among competing parameter values, so this model anticipates delays in parameter setting when critical input is sparse, and gradual setting of parameters. On the Structured Acquisition model, delays occur because parameters form a hierarchy, with higher-level parameters set before lower-level parameters. Assuming that children freely choose the initial value, children sometimes will miss-set parameters. However when that happens, the input is expected to trigger a precipitous rise in one parameter value and a corresponding decline in the other value. We will point to the kind of child language data that is needed in order to adjudicate among these competing models.

Is maximizing resilience compatible with established ecological goal functions?

Cropp and Gabric [Ecosystem adaptation: do ecosystems maximise resilience? Ecology. In press] used a simple phytoplanktonzooplankton-nutrient model and a genetic algorithm to determine the parameter values that would maximize the value of certain goal functions. These goal functions were to maximize biomass, maximize flux, maximize flux to biomass ratio, and maximize resilience. It was found that maximizing goal functions maximized resilience. The objective of this study was to investigate whether the Cropp and Gabric [Ecosystem adaptation: do ecosystems maximise resilience? Ecology. In press] result was indicative of a general ecosystem principle, or peculiar to the model and parameter ranges used. This study successfully replicated the Cropp and Gabric [Ecosystem adaptation: do ecosystems maximise resilience? Ecology. In press] experiment for a number of different model types, however, a different interpretation of the results is made. A new metric, concordance, was devised to describe the agreement between goal functions. It was found that resilience has the highest concordance of all goal functions trialled. for most model types. This implies that resilience offers a compromise between the established ecological goal functions. The parameter value range used is found to affect the parameter versus goal function relationships. Local maxima and minima affected the relationship between parameters and goal functions, and between goal functions. (C) 2003 Elsevier B.V. All rights reserved.

Pseudo-randomness of round-off errors in discretized linear maps on the plane

We analyze the sequences of round-off errors of the orbits of a discretized planar rotation, from a probabilistic angle. It was shown [Bosio & Vivaldi, 2000] that for a dense set of parameters, the discretized map can be embedded into an expanding p-adic dynamical system, which serves as a source of deterministic randomness. For each parameter value, these systems can generate infinitely many distinct pseudo-random sequences over a finite alphabet, whose average period is conjectured to grow exponentially with the bit-length of the initial condition (the seed). We study some properties of these symbolic sequences, deriving a central limit theorem for the deviations between round-off and exact orbits, and obtain bounds concerning repetitions of words. We also explore some asymptotic problems computationally, verifying, among other things, that the occurrence of words of a given length is consistent with that of an abstract Bernoulli sequence.

The cost of uniqueness in groundwater model calibration

Calibration of a groundwater model requires that hydraulic properties be estimated throughout a model domain. This generally constitutes an underdetermined inverse problem, for which a Solution can only be found when some kind of regularization device is included in the inversion process. Inclusion of regularization in the calibration process can be implicit, for example through the use of zones of constant parameter value, or explicit, for example through solution of a constrained minimization problem in which parameters are made to respect preferred values, or preferred relationships, to the degree necessary for a unique solution to be obtained. The cost of uniqueness is this: no matter which regularization methodology is employed, the inevitable consequence of its use is a loss of detail in the calibrated field. This, ill turn, can lead to erroneous predictions made by a model that is ostensibly well calibrated. Information made available as a by-product of the regularized inversion process allows the reasons for this loss of detail to be better understood. In particular, it is easily demonstrated that the estimated value for an hydraulic property at any point within a model domain is, in fact, a weighted average of the true hydraulic property over a much larger area. This averaging process causes loss of resolution in the estimated field. Where hydraulic conductivity is the hydraulic property being estimated, high averaging weights exist in areas that are strategically disposed with respect to measurement wells, while other areas may contribute very little to the estimated hydraulic conductivity at any point within the model domain, this possibly making the detection of hydraulic conductivity anomalies in these latter areas almost impossible. A study of the post-calibration parameter field covariance matrix allows further insights into the loss of system detail incurred through the calibration process to be gained. A comparison of pre- and post-calibration parameter covariance matrices shows that the latter often possess a much smaller spectral bandwidth than the former. It is also demonstrated that, as all inevitable consequence of the fact that a calibrated model cannot replicate every detail of the true system, model-to-measurement residuals can show a high degree of spatial correlation, a fact which must be taken into account when assessing these residuals either qualitatively, or quantitatively in the exploration of model predictive uncertainty. These principles are demonstrated using a synthetic case in which spatial parameter definition is based oil pilot points, and calibration is Implemented using both zones of piecewise constancy and constrained minimization regularization. (C) 2005 Elsevier Ltd. All rights reserved.

An empirical study of Hoelfding Racing for model selction in K-nearest neighbor classification

Racing algorithms have recently been proposed as a general-purpose method for performing model selection in machine teaming algorithms. In this paper, we present an empirical study of the Hoeffding racing algorithm for selecting the k parameter in a simple k-nearest neighbor classifier. Fifteen widely-used classification datasets from UCI are used and experiments conducted across different confidence levels for racing. The results reveal a significant amount of sensitivity of the k-nn classifier to its model parameter value. The Hoeffding racing algorithm also varies widely in its performance, in terms of the computational savings gained over an exhaustive evaluation. While in some cases the savings gained are quite small, the racing algorithm proved to be highly robust to the possibility of erroneously eliminating the optimal models. All results were strongly dependent on the datasets used.

Robust parameter settings of evolutionary algorithms for the optimisation of agricultural systems models

Numerical optimisation methods are being more commonly applied to agricultural systems models, to identify the most profitable management strategies. The available optimisation algorithms are reviewed and compared, with literature and our studies identifying evolutionary algorithms (including genetic algorithms) as superior in this regard to simulated annealing, tabu search, hill-climbing, and direct-search methods. Results of a complex beef property optimisation, using a real-value genetic algorithm, are presented. The relative contributions of the range of operational options and parameters of this method are discussed, and general recommendations listed to assist practitioners applying evolutionary algorithms to the solution of agricultural systems. (C) 2001 Elsevier Science Ltd. All rights reserved.

Effects of heat treatment on the nutritional value of raw soybean selected for low trypsin inhibitor activity

1. Two broiler experiments and a layer experiments were conducted on Kunitz trypsin inhibitor (Kti) soybeans (SB) of low trypsin inhibitor (TI) activity to determine their nutritive value when included as mash in least-cost poultry diets. 2. Experiment 1 compared chick performance on the Kti or raw SB using a commercial full-fat SB meal (FFSBM) and a solvent extracted SB meal (SBM) as controls during a 20 d experimental period. Broiler experiment 2 compared Kti and raw SB, non-steamed, or steam-pelleted with and without DL-methionine supplementation added to every treatment containing 170 g SB/kg. For each broiler experiment the levels of each SB were 70, 120 and 170 g/kg with the control birds fed only 170 g SB/kg. 3. The layer experiment, compared steam-pelleted Kti and raw SB against a non-steamed Kti and raw SB each fed at two levels (70 and 110 g/kg) x 30 replicates from 29 weeks of age for 19 weeks in a completely randomised design. Production parameters were measured when diets were formulated to contain minimum required specifications and calculated apparent metabolisable energy (AME). At the completion of each trial, 2 broiler birds from each cage and 5 layer birds per treatment were killed, weighed, and their liver and pancreas weighed. 4. Both broiler experiments indicated that production parameters on the Kti SB treatments were significantly lower (P < 0.05) than on the two commercial control SB treatments. However, the Kti treatments were superior to the raw SB treatments. 5. Pancreas weight increased with increasing inclusion of both raw and Kti SB, suggesting that a TI was causing the depression in performance. The AME of the Kti SB was similar to that of commercial FFSB meal. After steam conditioning, the raw SB meal AME value of 9.5 MJ/kg dry matter (DM) was improved to 14.1 MJ/kg DM by reduced TI activity, but this AME improvement with TI activity reduction, plus the supplementation with DL-methionine on birds fed the raw SB had no effect (P > 0.05) on any parameter evaluated in experiment 2. 6. The layer experiment showed that hens on the Kti SB treatments had significantly greater live weight gain (LWG), egg weight and daily egg mass than birds given raw SB. A reduced food intake (FI) was observed in the Kti treatments but egg mass was generally similar to that on the FFSB control diet, indicating that Kti SB supported excellent egg production at an inclusion of 110 g/kg. The depressed performance observed for broiler chicks suggest that younger birds are more susceptible to the effects of SB TI.

Global behavior of positive solutions of nonlinear three-point boundary value problems

We investigate the structure of the positive solution set for nonlinear three-point boundary value problems of the form u('') + h(t) f(u) = 0, u(0) = 0, u(1) = lambdau(eta), where eta epsilon (0, 1) is given lambda epsilon (0, 1/n) is a parameter, f epsilon C ([0, infinity), [0, infinity)) satisfies f (s) > 0 for s > 0, and h epsilon C([0, 1], [0, infinity)) is not identically zero on any subinterval of [0, 1]. Our main results demonstrate the existence of continua of positive solutions of the above problem. (C) 2004 Elsevier Ltd. All rights reserved.

Procedures and parameters in the real-time program refinement calculus

The real-time refinement calculus is a formal method for the systematic derivation of real-time programs from real-time specifications in a style similar to the non-real-time refinement calculi of Back and Morgan. In this paper we extend the real-time refinement calculus with procedures and provide refinement rules for refining real-time specifications to procedure calls. A real-time specification can include constraints on, not only what outputs are produced, but also when they are produced. The derived programs can also include time constraints oil when certain points in the program must be reached; these are expressed in the form of deadline commands. Such programs are machine independent. An important consequence of the approach taken is that, not only are the specifications machine independent, but the whole refinement process is machine independent. To implement the machine independent code on a target machine one has a separate task of showing that the compiled machine code will reach all its deadlines before they expire. For real-time programs, externally observable input and output variables are essential. These differ from local variables in that their values are observable over the duration of the execution of the program. Hence procedures require input and output parameter mechanisms that are references to the actual parameters so that changes to external inputs are observable within the procedure and changes to output parameters are externally observable. In addition, we allow value and result parameters. These may be auxiliary parameters, which are used for reasoning about the correctness of real-time programs as well as in the expression of timing deadlines, but do not lead to any code being generated for them by a compiler. (c) 2006 Elsevier B.V. All rights reserved.

Adult language input in the early childhood educational setting

Age-related changes in the adult language addressed to children aged 2;0-4;0 years in polyadic conditions were investigated in Australian childcare centres. The language that 21 staff members addressed to these children was coded for multiple variables in the broad social categories of prosody, context, speech act and gesture. The linguistic components were coded within the categories of phonology, lexicon, morphology, syntax and referential deixis. Minimal age-related differences were found. Explanations for the similarity of the adult language input across the age groups within the early childhood educational environment, will be discussed

The effects of Australian tall poppy attitudes on American value based leadership theory

A survey study of twenty-two Australian CEOs and their subordinates assessed relationships between Australian leader motives, Australian value based leader behaviour, subordinate tall poppy attitudes and subordinate commitment, effectiveness, motivation and satisfaction (CEMS). On the whole, the results showed general support for value based leadership processes. Subsequent regression analyses of the second main component of Value Based Leadership Theory, value based leader behaviour, revealed that the collectivistic, inspirational, integrity and visionary behaviour sub-scales of the construct were positively related with subordinate CEMS. Although the hypothesis that subordinate tall poppy attitudes would moderate value based leadership processes was not clearly supported, subsequent regression analyses found that subordinate tall poppy attitudes were negatively related with perceptions of value based leader behaviour and CEMS. These findings suggest complex relationships between the three constructs, and the proposed model for the Australian context is accordingly amended. Overall, the research supports the need to consider cultural-specific attitudes in management development.

A fast cross-entropy method for estimating buffer overflows in queuing networks

In this paper, we propose a fast adaptive importance sampling method for the efficient simulation of buffer overflow probabilities in queueing networks. The method comprises three stages. First, we estimate the minimum cross-entropy tilting parameter for a small buffer level; next, we use this as a starting value for the estimation of the optimal tilting parameter for the actual (large) buffer level. Finally, the tilting parameter just found is used to estimate the overflow probability of interest. We study various properties of the method in more detail for the M/M/1 queue and conjecture that similar properties also hold for quite general queueing networks. Numerical results support this conjecture and demonstrate the high efficiency of the proposed algorithm.

Existence of Multiple Solutions for Second Order Boundary Value Problems

We give conditions on f involving pairs of lower and upper solutions which lead to the existence of at least three solutions of the two point boundary value problem y" + f(x, y, y') = 0, x epsilon [0, 1], y(0) = 0 = y(1). In the special case f(x, y, y') = f(y) greater than or equal to 0 we give growth conditions on f and apply our general result to show the existence of three positive solutions. We give an example showing this latter result is sharp. Our results extend those of Avery and of Lakshmikantham et al.

Transient response of diffusion cell containing a porous solid

Transient response of an adsorbing or non-adsorbing tracer injected as step or square pulse input in a diffusion cell with two flowing streams across the pellet is theoretically investigated in this paper. Exact solutions and the asymptotic solutions in the time domain and in three different limits are obtained by using an integral transform technique and a singular perturbation technique, respectively. Parametric dependence of the concentrations in the top and bottom chambers can be revealed by investigating the asymptotic solutions, which are far simpler than their exact counterpart. In the time domain investigation, it is found that the bottom-chamber concentration is very sensitive to the value of the macropore effective diffusivity. Therefore this concentration could be used to extract diffusivity by fitting in the time domain. The bottom-chamber concentration is also sensitive to flow rate, pellet length chamber volume and the type of input (step and square input).

An improved novel method for solving nonlinear boundary value problems

A modified formula for the integral transform of a nonlinear function is proposed for a class of nonlinear boundary value problems. The technique presented in this paper results in analytical solutions. Iterations and initial guess, which are needed in other techniques, are not required in this novel technique. The analytical solutions are found to agree surprisingly well with the numerically exact solutions for two examples of power law reaction and Langmuir-Hinshelwood reaction in a catalyst pellet.