Two-Stage Orthogonal Least Squares Methods for Neural Network Construction

A number of neural networks can be formulated as the linear-in-the-parameters models. Training such networks can be transformed to a model selection problem where a compact model is selected from all the candidates using subset selection algorithms. Forward selection methods are popular fast subset selection approaches. However, they may only produce suboptimal models and can be trapped into a local minimum. More recently, a two-stage fast recursive algorithm (TSFRA) combining forward selection and backward model refinement has been proposed to improve the compactness and generalization performance of the model. This paper proposes unified two-stage orthogonal least squares methods instead of the fast recursive-based methods. In contrast to the TSFRA, this paper derives a new simplified relationship between the forward and the backward stages to avoid repetitive computations using the inherent orthogonal properties of the least squares methods. Furthermore, a new term exchanging scheme for backward model refinement is introduced to reduce computational demand. Finally, given the error reduction ratio criterion, effective and efficient forward and backward subset selection procedures are proposed. Extensive examples are presented to demonstrate the improved model compactness constructed by the proposed technique in comparison with some popular methods.

The role of complexity metrics in a multi-institutional dosimetry audit of VMAT

OBJECTIVE: To demonstrate the benefit of complexity metrics such as the modulation complexity score (MCS) and monitor units (MUs) in multi-institutional audits of volumetric-modulated arc therapy (VMAT) delivery.

METHODS: 39 VMAT treatment plans were analysed using MCS and MU. A virtual phantom planning exercise was planned and independently measured using the PTW Octavius(®) phantom and seven29(®) 2D array (PTW-Freiburg GmbH, Freiburg, Germany). MCS and MU were compared with the median gamma index pass rates (2%/2 and 3%/3 mm) and plan quality. The treatment planning systems (TPS) were grouped by VMAT modelling being specifically designed for the linear accelerator manufacturer's own treatment delivery system (Type 1) or independent of vendor for VMAT delivery (Type 2). Differences in plan complexity (MCS and MU) between TPS types were compared.

RESULTS: For Varian(®) linear accelerators (Varian(®) Medical Systems, Inc., Palo Alto, CA), MCS and MU were significantly correlated with gamma pass rates. Type 2 TPS created poorer quality, more complex plans with significantly higher MUs and MCS than Type 1 TPS. Plan quality was significantly correlated with MU for Type 2 plans. A statistically significant correlation was observed between MU and MCS for all plans (R = -0.84, p < 0.01).

CONCLUSION: MU and MCS have a role in assessing plan complexity in audits along with plan quality metrics. Plan complexity metrics give some indication of plan deliverability but should be analysed with plan quality.

ADVANCES IN KNOWLEDGE: Complexity metrics were investigated for a national rotational audit involving 34 institutions and they showed value. The metrics found that more complex plans were created for planning systems which were independent of vendor for VMAT delivery.

Corneal hysteresis and axial length among Chinese secondary school children: the Xichang Pediatric Refractive Error Study (X-PRES) report no. 4.

PURPOSE: To evaluate the association between corneal hysteresis and axial length/refractive error among rural Chinese secondary school children. DESIGN: Cross-sectional cohort study. METHODS: Refractive error (cycloplegic auto-refraction with subjective refinement), central corneal thickness (CCT) and axial length (ultrasonic measurement), intraocular pressure (IOP), and corneal hysteresis (Reichert Ocular Response Analyzer) were measured on a rural school-based cohort of children. RESULTS: Among 1,233 examined children, the mean age was 14.7 +/- 0.8 years and 699 (56.7%) were girls. The mean spherical equivalent (n = 1,232) was -2.2 +/- 1.6 diopters (D), axial length (n = 643) was 23.7 +/- 1.1 mm, corneal hysteresis (n = 1,153) was 10.7 +/- 1.6 mm Hg, IOP (n = 1,153) was 17.0 +/- 3.4 mm Hg, and CCT (n = 1,226) was 553 +/- 33 microns. In linear regression models, longer axial length was significantly (P < .001 for both) associated with lower corneal hysteresis and higher IOP. Hysteresis in this population was significantly (P < .001) lower than has previously been reported for normal White children (n = 42, 12.3 +/- 1.3 mm Hg), when adjusting for age and gender. This difference did not appear to depend on differences in axial length between the populations, as it persists when only Chinese children with normal uncorrected vision are included. CONCLUSIONS: Prospective studies will be needed to determine if low hysteresis places eyes at risk for axial elongation secondary or if primary elongation results in lower hysteresis.

A Cryptosystem Using the Concepts of Algebraic Geometric Code

Cryptosystem using linear codes was developed in 1978 by Mc-Eliece. Later in 1985 Niederreiter and others developed a modified version of cryptosystem using concepts of linear codes. But these systems were not used frequently because of its larger key size. In this study we were designing a cryptosystem using the concepts of algebraic geometric codes with smaller key size. Error detection and correction can be done efficiently by simple decoding methods using the cryptosystem developed. Approach: Algebraic geometric codes are codes, generated using curves. The cryptosystem use basic concepts of elliptic curves cryptography and generator matrix. Decrypted information takes the form of a repetition code. Due to this complexity of decoding procedure is reduced. Error detection and correction can be carried out efficiently by solving a simple system of linear equations, there by imposing the concepts of security along with error detection and correction. Results: Implementation of the algorithm is done on MATLAB and comparative analysis is also done on various parameters of the system. Attacks are common to all cryptosystems. But by securely choosing curve, field and representation of elements in field, we can overcome the attacks and a stable system can be generated. Conclusion: The algorithm defined here protects the information from an intruder and also from the error in communication channel by efficient error correction methods.

Polyphase Implementation of Non-recursive Comb Decimators for Sigma-Delta A/D Converters

In a sigma-delta analog to digital (A/D) As most of the sigma-delta ADC applications require converter, the most computationally intensive block is decimation filters with linear phase characteristics, the decimation filter and its hardware implementation symmetric Finite Impulse Response (FIR) filters are may require millions of transistors. Since these widely used for implementation. But the number of FIR converters are now targeted for a portable application, filter coefficients will be quite large for implementing a a hardware efficient design is an implicit requirement. narrow band decimation filter. Implementing decimation In this effect, this paper presents a computationally filter in several stages reduces the total number of filter efficient polyphase implementation of non-recursive coefficients, and hence reduces the hardware complexity cascaded integrator comb (CIC) decimators for and power consumption [2]. Sigma-Delta Converters (SDCs). The SDCs are The first stage of decimation filter can be operating at high oversampling frequencies and hence implemented very efficiently using a cascade of integrators require large sampling rate conversions. The filtering and comb filters which do not require multiplication or and rate reduction are performed in several stages to coefficient storage. The remaining filtering is performed reduce hardware complexity and power dissipation. either in single stage or in two stages with more complex The CIC filters are widely adopted as the first stage of FIR or infinite impulse response (IIR) filters according to decimation due to its multiplier free structure. In this the requirements. The amount of passband aliasing or research, the performance of polyphase structure is imaging error can be brought within prescribed bounds by compared with the CICs using recursive and increasing the number of stages in the CIC filter. The non-recursive algorithms in terms of power, speed and width of the passband and the frequency characteristics area. This polyphase implementation offers high speed outside the passband are severely limited. So, CIC filters operation and low power consumption. The polyphase are used to make the transition between high and low implementation of 4th order CIC filter with a sampling rates. Conventional filters operating at low decimation factor of '64' and input word length of sampling rate are used to attain the required transition '4-bits' offers about 70% and 37% of power saving bandwidth and stopband attenuation. compared to the corresponding recursive and Several papers are available in literature that deals non-recursive implementations respectively. The same with different implementations of decimation filter polyphase CIC filter can operate about 7 times faster architecture for sigma-delta ADCs. Hogenauer has than the recursive and about 3.7 times faster than the described the design procedures for decimation and non-recursive CIC filters.

Certified Rapid Solution of Parametrized Linear Elliptic Equations: Application to Parameter Estimation

We present a technique for the rapid and reliable evaluation of linear-functional output of elliptic partial differential equations with affine parameter dependence. The essential components are (i) rapidly uniformly convergent reduced-basis approximations — Galerkin projection onto a space WN spanned by solutions of the governing partial differential equation at N (optimally) selected points in parameter space; (ii) a posteriori error estimation — relaxations of the residual equation that provide inexpensive yet sharp and rigorous bounds for the error in the outputs; and (iii) offline/online computational procedures — stratagems that exploit affine parameter dependence to de-couple the generation and projection stages of the approximation process. The operation count for the online stage — in which, given a new parameter value, we calculate the output and associated error bound — depends only on N (typically small) and the parametric complexity of the problem. The method is thus ideally suited to the many-query and real-time contexts. In this paper, based on the technique we develop a robust inverse computational method for very fast solution of inverse problems characterized by parametrized partial differential equations. The essential ideas are in three-fold: first, we apply the technique to the forward problem for the rapid certified evaluation of PDE input-output relations and associated rigorous error bounds; second, we incorporate the reduced-basis approximation and error bounds into the inverse problem formulation; and third, rather than regularize the goodness-of-fit objective, we may instead identify all (or almost all, in the probabilistic sense) system configurations consistent with the available experimental data — well-posedness is reflected in a bounded "possibility region" that furthermore shrinks as the experimental error is decreased.

Forecasting the spot prices of various coffee types using linear and non-linear error correction models

Este documento estima modelos lineales y no-lineales de corrección de errores para los precios spot de cuatro tipos de café. En concordancia con las leyes económicas, se encuentra evidencia que cuando los precios están por encima de su nivel de equilibrio, retornan a éste mas lentamente que cuando están por debajo. Esto puede reflejar el hecho que, en el corto plazo, para los países productores de café es mas fácil restringir la oferta para incrementar precios, que incrementarla para reducirlos. Además, se encuentra evidencia que el ajuste es más rápido cuando las desviaciones del equilibrio son mayores. Los pronósticos que se obtienen a partir de los modelos de corrección de errores no lineales y asimétricos considerados en el trabajo, ofrecen una leve mejoría cuando se comparan con los pronósticos que resultan de un modelo de paseo aleatorio.

Selected configuration interaction with truncation energy error and application to the Ne atom

Selected configuration interaction (SCI) for atomic and molecular electronic structure calculations is reformulated in a general framework encompassing all CI methods. The linked cluster expansion is used as an intermediate device to approximate CI coefficients BK of disconnected configurations (those that can be expressed as products of combinations of singly and doubly excited ones) in terms of CI coefficients of lower-excited configurations where each K is a linear combination of configuration-state-functions (CSFs) over all degenerate elements of K. Disconnected configurations up to sextuply excited ones are selected by Brown's energy formula, ΔEK=(E-HKK)BK2/(1-BK2), with BK determined from coefficients of singly and doubly excited configurations. The truncation energy error from disconnected configurations, Δdis, is approximated by the sum of ΔEKS of all discarded Ks. The remaining (connected) configurations are selected by thresholds based on natural orbital concepts. Given a model CI space M, a usual upper bound ES is computed by CI in a selected space S, and EM=E S+ΔEdis+δE, where δE is a residual error which can be calculated by well-defined sensitivity analyses. An SCI calculation on Ne ground state featuring 1077 orbitals is presented. Convergence to within near spectroscopic accuracy (0.5 cm-1) is achieved in a model space M of 1.4× 109 CSFs (1.1 × 1012 determinants) containing up to quadruply excited CSFs. Accurate energy contributions of quintuples and sextuples in a model space of 6.5 × 1012 CSFs are obtained. The impact of SCI on various orbital methods is discussed. Since ΔEdis can readily be calculated for very large basis sets without the need of a CI calculation, it can be used to estimate the orbital basis incompleteness error. A method for precise and efficient evaluation of ES is taken up in a companion paper

Bayesian estimation and selection of nonlinear vector error correction models: The case of the sugar-ethanol-oil nexus in Brazil

Nonlinear adjustment toward long-run price equilibrium relationships in the sugar-ethanol-oil nexus in Brazil is examined. We develop generalized bivariate error correction models that allow for cointegration between sugar, ethanol, and oil prices, where dynamic adjustments are potentially nonlinear functions of the disequilibrium errors. A range of models are estimated using Bayesian Monte Carlo Markov Chain algorithms and compared using Bayesian model selection methods. The results suggest that the long-run drivers of Brazilian sugar prices are oil prices and that there are nonlinearities in the adjustment processes of sugar and ethanol prices to oil price but linear adjustment between ethanol and sugar prices.

Design of a minimum variance multiple input-multiple output neuro self-tuning proportional-integral-derivative controller for non-linear dynamic systems

In this study a minimum variance neuro self-tuning proportional-integral-derivative (PID) controller is designed for complex multiple input-multiple output (MIMO) dynamic systems. An approximation model is constructed, which consists of two functional blocks. The first block uses a linear submodel to approximate dominant system dynamics around a selected number of operating points. The second block is used as an error agent, implemented by a neural network, to accommodate the inaccuracy possibly introduced by the linear submodel approximation, various complexities/uncertainties, and complicated coupling effects frequently exhibited in non-linear MIMO dynamic systems. With the proposed model structure, controller design of an MIMO plant with n inputs and n outputs could be, for example, decomposed into n independent single input-single output (SISO) subsystem designs. The effectiveness of the controller design procedure is initially verified through simulations of industrial examples.

Linear-wavelet models for system identification

A model structure comprising a wavelet network and a linear term is proposed for nonlinear system identification. It is shown that under certain conditions wavelets are orthogonal to linear functions and, as a result, the two parts of the model can be identified separately. The linear-wavelet model is compared to a standard wavelet network using data from a simulated fermentation process. The results show that the linear-wavelet model yields a smaller modelling error when compared to a wavelet network using the same number of regressors.

Linear-wavelet networks

This paper proposes a nonlinear regression structure comprising a wavelet network and a linear term. The introduction of the linear term is aimed at providing a more parsimonious interpolation in high-dimensional spaces when the modelling samples are sparse. A constructive procedure for building such structures, termed linear-wavelet networks, is described. For illustration, the proposed procedure is employed in the framework of dynamic system identification. In an example involving a simulated fermentation process, it is shown that a linear-wavelet network yields a smaller approximation error when compared with a wavelet network with the same number of regressors. The proposed technique is also applied to the identification of a pressure plant from experimental data. In this case, the results show that the introduction of wavelets considerably improves the prediction ability of a linear model. Standard errors on the estimated model coefficients are also calculated to assess the numerical conditioning of the identification process.

Cyclic prefixed OQAM-OFDM and its application to single-carrier FDMA

Single-carrier frequency division multiple access (SC-FDMA) has appeared to be a promising technique for high data rate uplink communications. Aimed at SC-FDMA applications, a cyclic prefixed version of the offset quadrature amplitude modulation based OFDM (OQAM-OFDM) is first proposed in this paper. We show that cyclic prefixed OQAMOFDM CP-OQAM-OFDM) can be realized within the framework of the standard OFDM system, and perfect recovery condition in the ideal channel is derived. We then apply CP-OQAMOFDM to SC-FDMA transmission in frequency selective fading channels. Signal model and joint widely linear minimum mean square error (WLMMSE) equalization using a prior information with low complexity are developed. Compared with the existing DFTS-OFDM based SC-FDMA, the proposed SC-FDMA can significantly reduce envelope fluctuation (EF) of the transmitted signal while maintaining the bandwidth efficiency. The inherent structure of CP-OQAM-OFDM enables low-complexity joint equalization in the frequency domain to combat both the multiple access interference and the intersymbol interference. The joint WLMMSE equalization using a prior information guarantees optimal MMSE performance and supports Turbo receiver for improved bit error rate (BER) performance. Simulation resultsconfirm the effectiveness of the proposed SC-FDMA in termsof EF (including peak-to-average power ratio, instantaneous-toaverage power ratio and cubic metric) and BER performances.

Sparse probability density function estimation using the minimum integrated square error

We develop a new sparse kernel density estimator using a forward constrained regression framework, within which the nonnegative and summing-to-unity constraints of the mixing weights can easily be satisfied. Our main contribution is to derive a recursive algorithm to select significant kernels one at time based on the minimum integrated square error (MISE) criterion for both the selection of kernels and the estimation of mixing weights. The proposed approach is simple to implement and the associated computational cost is very low. Specifically, the complexity of our algorithm is in the order of the number of training data N, which is much lower than the order of N2 offered by the best existing sparse kernel density estimators. Numerical examples are employed to demonstrate that the proposed approach is effective in constructing sparse kernel density estimators with comparable accuracy to those of the classical Parzen window estimate and other existing sparse kernel density estimators.