Model Selection Scheme Research Articles

An identification technique for data acquisition characterization in the presence of nonlinear distortions and time base distortions

The nonlinear behavior of data acquisition channels and analog-to-digital converters is often measured using sine-wave measurements. High-frequency sampling scopes also suffer from time base distortions. This implies that the signals are sampled at a nonequidistant time grid. This paper describes a robust and efficient identification technique to characterize acquisition channels which suffer from both nonlinear distortions and/or time base distortions in the presence of both additive and jitter noise. An automatic model selection scheme and the generation of uncertainty bounds are obtained through the statistical properties of the proposed simulator. The applicability of the method is demonstrated on both simulations and measurements of high-frequency sampling scopes.

Statistical model selection based on Bayes decision theory and its application to change detection problem

The statistical model selection problems are discussed, and the criterion based on the Bayes decision theory for the detection of the true model is derived. We propose a new Bayesian model selection scheme, which can detect the change points of the parameter of the information source and also estimate each unknown parameter. In this paper, we formulate the Bayes optimal solution of the change detection problem, and analyze its property of the consistency.

Predictive minimum description length criterion for time series modeling with neural networks.

Nonlinear time series modeling with a multilayer perceptron network is presented. An important aspect of this modeling is the model selection, i.e., the problem of determining the size as well as the complexity of the model. To overcome this problem we apply the predictive minimum description length (PMDL) principle as a minimization criterion. In the neural network scheme it means minimizing the number of input and hidden units. Three time series modeling experiments are used to examine the usefulness of the PMDL model selection scheme. A comparison with the widely used cross-validation technique is also presented. In our experiments the PMDL scheme and the cross-validation scheme yield similar results in terms of model complexity. However, the PMDL method was found to be two times faster to compute. This is significant improvement since model selection in general is very time consuming.

Sensitivity of Critical Load Estimates for Surface Waters to Model Selection and Regionalization Schemes

AbstractA critical load is the amount of atmospheric pollutant that can be deposited on a sensitive ecosystem without causing measurable, longterm degradation in ecosystem form or function. We compare several methods for making critical load estimates of SO4 deposition to lakes; these methods are among those identified by participants in the United Nations‐Economic Commission for Europe (UNECE) Convention on Long‐Range Transboundary Air Pollution. The models were originally developed to address surface water acidification issues. Critical loads are computed here as the levels of deposition required to reduce surface water acid neutralizing capacities (ANC) from present values to either 25 µeq L−1 or 0 µeq L−1. These end points were selected to provide a common basis for comparing critical load values obtained from the various models, and are not intended to provide definitive critical load estimates. Using the amount of deposition required to decrease steady‐state surface water ANC values from 25 to 0 µeq L−1, about 9 ± 2 kgSO4 ha−1 yr−1 in the northeastern USA as a basis for comparison, we find that models make significantly different projections regarding ecosystem sensitivites to deposition. For lakes with critical load values in the range of current deposition in the Northeast (i.e., 18–56 kg SO4 ha−1 yr−1), model results differ by as much as 82 kg SO4 ha−1 yr−1. Differences in results are observed not only between models, but within models when slightly different approaches are used for estimating certain parameters. The interpretation of model results depends strongly on how systems are grouped regionally. While the results suggest that reductions in deposition from current levels would reduce adverse effects in sensitive lakes, the magnitude and timing of the reductions necessary and the determination of appropriate population end points are decisions that will ultimately be made by risk managers rather than risk assessors.