Backward elimination model construction for regression and classification using leave-one-out criteria

A fundamental principle in practical nonlinear data modeling is the parsimonious principle of constructing the minimal model that explains the training data well. Leave-one-out (LOO) cross validation is often used to estimate generalization errors by choosing amongst different network architectures (M. Stone, "Cross validatory choice and assessment of statistical predictions", J. R. Stast. Soc., Ser. B, 36, pp. 117-147, 1974). Based upon the minimization of LOO criteria of either the mean squares of LOO errors or the LOO misclassification rate respectively, we present two backward elimination algorithms as model post-processing procedures for regression and classification problems. The proposed backward elimination procedures exploit an orthogonalization procedure to enable the orthogonality between the subspace as spanned by the pruned model and the deleted regressor. Subsequently, it is shown that the LOO criteria used in both algorithms can be calculated via some analytic recursive formula, as derived in this contribution, without actually splitting the estimation data set so as to reduce computational expense. Compared to most other model construction methods, the proposed algorithms are advantageous in several aspects; (i) There are no tuning parameters to be optimized through an extra validation data set; (ii) The procedure is fully automatic without an additional stopping criteria; and (iii) The model structure selection is directly based on model generalization performance. The illustrative examples on regression and classification are used to demonstrate that the proposed algorithms are viable post-processing methods to prune a model to gain extra sparsity and improved generalization.

Givens rotation based fast backward elimination algorithm for RBF neural network pruning

A fast backward elimination algorithm is introduced based on a QR decomposition and Givens transformations to prune radial-basis-function networks. Nodes are sequentially removed using an increment of error variance criterion. The procedure is terminated by using a prediction risk criterion so as to obtain a model structure with good generalisation properties. The algorithm can be used to postprocess radial basis centres selected using a k-means routine and, in this mode, it provides a hybrid supervised centre selection approach.

Systematic selection of analogue redesign method for forward-type digital power converters

This article proposes a systematic approach to determine the most suitable analogue redesign method to be used for forward-type converters under digital voltage mode control. The focus of the method is to achieve the highest phase margin at the particular switching and crossover frequencies chosen by the designer. It is shown that at high crossover frequencies with respect to switching frequency, controllers designed using backward integration have the largest phase margin; whereas at low crossover frequencies with respect to switching frequency, controllers designed using bilinear integration with pre-warping have the largest phase margins. An algorithm has been developed to determine the frequency of the crossing point where the recommended discretisation method changes. An accurate model of the power stage is used for simulation and experimental results from a Buck converter are collected. The performance of the digital controllers is compared to that of the equivalent analogue controller both in simulation and experiment. Excellent closeness between the simulation and experimental results is presented. This work provides a concrete example to allow academics and engineers to systematically choose a discretisation method.

Wavelet-based dynamic time warping

Dynamic Time Warping (DTW), a pattern matching technique traditionally used for restricted vocabulary speech recognition, is based on a temporal alignment of the input signal with the template models. The principal drawback of DTW is its high computational cost as the lengths of the signals increase. This paper shows extended results over our previously published conference paper, which introduces an optimized version of the DTW I hat is based on the Discrete Wavelet Transform (DWT). (C) 2008 Elsevier B.V. All rights reserved.

Nash equilibrium under knightian uncertainty: breaking down backward induction (extensively revised version)

We define Nash equilibrium for two-person normal form games in the presence of uncertainty, in the sense of Knight(1921). We use the fonna1iution of uncertainty due to Schmeidler and Gilboa. We show tbat there exist Nash equilibria for any degree of uncertainty, as measured by the uncertainty aversion (Dow anel Wer1ang(l992a». We show by example tbat prudent behaviour (maxmin) can be obtained as an outcome even when it is not rationaliuble in the usual sense. Next, we break down backward industion in the twice repeated prisoner's dilemma. We link these results with those on cooperation in the finitely repeated prisoner's dilemma obtained by Kreps-Milgrom-Roberts-Wdson(1982), and withthe 1iterature on epistemological conditions underlying Nash equilibrium. The knowledge notion implicit in this mode1 of equilibrium does not display logical omniscience.

Nash equilibrium under knightian uncertainty: breaking-down backward induction

We present two alternative definitions of Nash equilibrium for two person games in the presence af uncertainty, in the sense of Knight. We use the formalization of uncertainty due to Schmeidler and Gilboa. We show that, with one of the definitions, prudent behaviour (maxmin) can be obtained as an outcome even when it is not rationalizable in the usual sense. Most striking is that with the Same definition we break down backward induction in the twice repeated prisoner's dilemma. We also link these results with the Kreps-Milgrom-Roberts-Wilson explanation of cooperation in the finitely repeated prisoner's dilemma.

Fast-forward/fast-backward substitutions on vector computers

This paper deals with approaches for sparse matrix substitutions using vector processing. Many publications have used the W-matrix method to solve the forward/backward substitutions on vector computer. Recently a different approach has been presented using dependency-based substitution algorithm (DBSA). In this paper the focus is on new algorithms able to explore the sparsity of the vectors. The efficiency is tested using linear systems from power systems with 118, 320, 725 and 1729 buses. The tests were performed on a CRAY Y MP2E/232. The speedups for a fast-forward/fast-backward using a 1729-bus system are near 19 and 14 for real and complex arithmetic operations, respectively. When forward/backward is employed the speedups are about 8 and 6 to perform the same simulations.

Measurement of the forward-backward charge asymmetry in top-quark pair production

We present the first measurement of the integrated forward-backward charge asymmetry in top-quark-top-antiquark pair (t (t) over bar) production in proton-antiproton (p (p) over bar) collisions in the lepton+jets final state. Using a b-jet tagging algorithm and kinematic reconstruction assuming t (t) over bar +X production and decay, a sample of 0.9 fb(-1) of data, collected by the D0 experiment at the Fermilab Tevatron Collider, is used to measure the asymmetry for different jet multiplicities. The result is also used to set upper limits on t (t) over bar +X production via a Z' resonance.

Measurement of the Forward-Backward Charge Asymmetry and Extraction of sin(2)theta(eff)(W) in pp -> Z/gamma(*)+X -> e(+)e(-)+X Events Produced at s=1.96 TeV

We present a measurement of the forward-backward charge asymmetry (A(FB)) in pp -> Z/gamma(*)+X -> e(+)e(-)+X events at a center-of-mass energy of 1.96 TeV using 1.1 fb(-1) of data collected with the D0 detector at the Fermilab Tevatron collider. A(FB) is measured as a function of the invariant mass of the electron-positron pair, and found to be consistent with the standard model prediction. We use the A(FB) measurement to extract the effective weak mixing angle sin(2)theta(eff)(W)=0.2326 +/- 0.0018(stat)+/- 0.0006(syst).

Measurement of sin(2)theta(l)(eff) and Z-light quark couplings using the forward-backward charge asymmetry in p(p)over-bar -> Z/gamma* -> e(+)e(-) events with L=5.0 fb(-1) at root s=1.96 TeV

Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)