Adaptive Least-squares Research Articles

Modeling of Complex Manufacturing Processes by Hierarchical Fuzzy Basis Function Networks With Application to Grinding Processes

A framework for modeling complex manufacturing processes using fuzzy neural networks is presented with a novel training algorithm. In this study, a hierarchical structure that consists of fuzzy basis function networks (FBFN) is proposed to construct comprehensive models of the complex processes. A new adaptive least-squares (ALS) algorithm, based on the least-squares method and genetic algorithm (GA), is proposed for autonomous learning and construction of FBFNs without any human intervention. Simulation studies are performed to demonstrate advantages of the proposed modeling framework with the training algorithm in modeling complex manufacturing processes. The proposed method is implemented for the surface grinding processes based on the hierarchical structure of FBFNs. Process models for surface roughness and residual stress are developed based on the available grinding model structures with a small number of experimental data to demonstrate the concept. The accuracy of developed models is validated through independent sets of grinding experiments.

Decision feedback differential detection for differential orthogonal space-time modulation with APSK signals over flat-fading channels

Differential orthogonal space-time modulation (DOSTM) with amplitude/phase shift keying (APSK) signals has been recently proposed to improve the data throughput of the DOSTM with PSK signals over quasi-static channels. In this letter, decision feedback differential detection (DF-DD) based on linear prediction is presented for the DOSTM with APSK Signals (DOSTM-APSK) over flat-fading channels. The proposed DF-DD offers better performance than the differential detection when the channel experiences fast fading. The coefficients of the linear prediction based DF-DD can be obtained by an adaptive recursive least-squares algorithm, where the channel statistics are not needed. The proposed DF-DD is also applicable to the general unitary differential space-time modulation.

Adaptive Iterative Multiuser Decision Feedback Detection

Adaptive iterative receivers which combine multiuser decision-feedback detection with maximum a posteriori (MAP) decoding and soft feedback are presented for synchronous coded direct sequence-code-division multiple access. Both successive and parallel demodulation of users are considered. Optimal filters are derived using both minimum mean squared error and least squares (LS) criteria. The latter assumes short (repeated) spreading codes and that the users to be demodulated simultaneously transmit training sequences. The LS criterion does not require prior knowledge or estimates of spreading codes and channels. Simulation results show that the adaptive receiver can perform significantly better than the standard (soft) interference canceller, since the adaptive algorithm attempts to measure and exploit the second-order statistics between the input and output of the MAP decoder. With limited training, successive feedback and decoding performs significantly better than parallel feedback. The effect of code rate on performance is examined, and reduced-rank versions of the adaptive LS algorithms, which can reduce training overhead, are also presented.

Linear parameter estimation and predictive constrained control of wiener/hammerstein systems

A new, analytical, orthonormal basis functions (OBF)-based design methodology for adaptive predictive constrained control of open-loop stable, possibly nonminimum phase, time-varying Wiener and Hammerstein systems is presented. A linear adaptive least-squares parameter estimation algorithm is applied both to a nonlinear static part and a linear dynamic, OBF-modeled factor of the Wiener/Hammerstein system. A notion of inverse systems is crucial for linear estimation of both Wiener and Hammerstein systems, with in verses of the nonlinear or linear parts respectively involved. The adaptive estimator is coupled with a simple but robust, predictive control strategy called Extended Horizon Model Algorithmic Control, with input/output constraints handled in a trivial way. Simulation examples demonstrate computational and numerical effectiveness of the new adaptive nonlinear constrained control approach.

Construction of fuzzy systems using least-squares method and genetic algorithm

The fuzzy basis function network which was proposed in Wang and Mendel (IEEE Trans. Neural Networks 3(5) (1992b) 807) provides a way of representing fuzzy inference systems in a simple structure similar to those of radial basis function networks. In this paper, two new algorithms based on the least-squares method and genetic algorithm are proposed for autonomous learning and construction of fuzzy basis function networks when training data are available. The proposed algorithms add a significant fuzzy basis function node at each iteration during training, based on error reduction measures. The first, a least-squares algorithm, provides a way of sequentially constructing meaningful fuzzy systems which are not possible to achieve with the orthogonal least-squares algorithm, while the second, an adaptive least-squares algorithm based on the combined least-squares and genetic algorithm, realizes hybrid structure-parameter learning without human intervention. Simulation studies are performed with numerical examples for comparison of its performance against the orthogonal least-squares algorithm, backpropagation algorithm, and conventional genetic algorithm. The adaptive least-squares algorithm is also applied to a real world problem to construct a fuzzy basis function network model for surface roughness in a grinding process using experimental data.

Monetary Policy Rules with Model and Data Uncertainty

In this paper we examine the prevalence of data, specification, and parameter uncertainty in the formation of simple rules that mimic monetary policymaking decisions. Our approach is to build real-time data sets and simulate a real-time policy-setting environment in which we assume that policy is captured by movements in the actual federal funds rate, and then to assess what sorts of policy rule models and what sorts of data best explain what the Federal Reserve actually did. This approach allows us not only to track the performance of alternative rules over time (hence facilitating a type of model selection among competing rules), but also to more generally assess the importance of the data revision process in the formation of macroeconomic time series models. From the perspective of real-time data, our results suggest that the use of data that are erroneous, in the sense that they were not available at the time decisions could have been made based on forecasts from the rules, can lead to the selection of quantitatively different models. From the perspective of finding a rule that best approximates what the Federal Reserve Board (Fed) has actually done (and hence from the perspective of finding a rule that best approximates what the Fed will do in the future), we find that (i) our version of calibration is better than naive estimation, although both are dominated by an approach to rule formation based on the use of adaptive least-squares learning; (ii) rules based on data that are not seasonally adjusted are more reliable than those based on seasonally adjusted data; and (iii) rules based solely on preliminary data do not minimize mean square forecast error risk. In particular, early releases of data can be noisy, and for this reason it is useful to also use data that have been revised when making decisions using policy rules.

최소 제곱 무요소법을 이용한 적응 기법에 관한 연구

An h-adaptive scheme of first-order least-squares meshfree method is presented. A posteriori error estimates, which can be readily computed from the residual, are also presented. For elliptic problem the error indicators are further improved by applying the Aubin-Nitsche method. In the proposed refinement scheme, Voronoi cells are utilized to insert nodes at appropriate positions. Through numerical examples, it is demonstrated that the error indicators reveal good correlations with the actual errors and the adaptive first-order least-squares meshfree method is effectively applied to the localized problems such as the shock formation in fluid dynamics.

Evolutionary modelling and optimization of grinding processes

The objective of this study is to develop a framework of modelling the complex grinding processes and finding optimal process conditions to meet the general class of process requirements. In order to achieve the above goal, novel modelling schemes and optimization methods based on evolutionary algorithms (EA) are developed. The optimization problem of grinding processes can be formulated as a constrained non-linear programming problem with mixed-discrete variables. The adaptive least-squares (ALS) algorithm proposed by Lee and Shin's 1998 study is extended for modelling multi-input-multi-output (MIMO) complex grinding processes using fuzzy basis function networks (FBFN), while the modified evolution strategies (ES) is proposed for successful optimization of grinding processes. Two grinding optimization problems demonstrate the superior performance of the proposed scheme.

An adaptive least-squares method for the compressible Euler equations

An adaptive least-squares finite element method is used to solve the compressible Euler equations in two dimensions. Since the method is naturally diffusive, no explicit artificial viscosity is added to the formulation. The inherent artificial viscosity, however, is usually large and hence does not allow sharp resolution of discontinuities unless extremely fine grids are used. To remedy this, while retaining the advantages of the least-squares method, a moving-node grid adaptation technique is used. The outstanding feature of the adaptive method is its sensitivity to directional features like shock waves, leading to the automatic construction of adapted grids where the element edge(s) are strongly aligned with such flow phenomena. Using well-known transonic and supersonic test cases, it has been demonstrated that by coupling the least-squares method with a robust adaptive method shocks can be captured with high resolution despite using relatively coarse grids. Copyright © 1999 John Wiley & Sons, Ltd.

A direction set based algorithm for least squares problems in adaptive signal processing

A fast algorithm has been developed for solving a class of least squares problems arising in adaptive signal processing. The algorithm is based on the Powell and Zangwill direction set method for unconstrained minimization problems and is designed so as to fully take advantage of the special structure of the adaptive least squares (ALS) problems. It is a fast algorithm because it requires only O( N) multiplications for each system update where N is the dimension of the problem. The algorithm has been implemented and applied to applications in signal processing. Computer simulation results have shown that the algorithm is stable and converges fast often with a rate which is comparable to that of the known CG and RLS algorithm. The convergence behavior of the algorithm is studied in detail in this paper. The algorithm is shown to converge linearly for adaptive least squares problems in general and its rate of convergence can be faster than linear for some applications. A computational procedure is also designed to generate a set of near conjugate directions as initial directions to accelerate the convergence.

Towards Robust Adaptive Least-Squares Parameter Estimation with Internal Feedback

The new concepts of the ‘covariance matrix normalization’ and the ‘cascade’ structure of the adaptive least-squares estimator are shown to generalize and extend the use of internal information feedback in various robustness/alertness-oriented modifications to the standard ALS estimation algorithm. In the cascade estimation structure it is possible to ‘naturally’ stabilize, rather than maximize, the information matrix so that the covariance windup and blowup are effectively eliminated and the celebrated square root update of the covariance matrix is no longer needed Consequently, a new, ‘single-loop/cascade’ ALS MIMO estimation algorithm, enabling to effectively track both slow and jump parameter variations, is presented. The algorithm is coupled with a new, simple but robust predictive control scheme, offering a new adaptive MIMO control strategy.

Spontaneous otoacoustic emission frequency is modulated by heartbeat.

Detailed analysis of spontaneous otoacoustic emissions (SOAEs) in human subjects revealed that all stable SOAEs sufficiently above the noise floor to permit appropriate analysis have sidebands at multiples of approximately 1 Hz. This is consistent with the hypothesis that SOAEs are modulated by heartbeat. Simultaneous measurement of the rate of blood flow to the thumb and the separation of the spectral sidebands demonstrated that they covary (r = 0.982, p < 5 x 10(-10)). An adaptive least-squares fit (LSF) paradigm was developed to facilitate the measurement of the instantaneous frequency and amplitude of the signals. A combination of traditional spectral analyses and new LSF analyses showed that the sideband generation stems from frequency modulation of the emissions. If there is any amplitude modulation correlated with the blood flow, it is below the noise floor of the analysis. The frequency of the emission was at a minimum when the blood flow was maximal. Examination of alternative mechanisms using computer simulations suggests that these changes stem from changes of 10-100 ppm in the mass of the basilar membrane.

Extended Horizon Adaptive Model Algorithmic Control

A new, original, robust adaptive control strategy termed Extended Horizon Adaptive Model Algorithmic Control is presented. In EHAMAC, a new, combined, ’single-loop’/’cascade’ adaptive least-squares parameter estimator is coupled with a new, simple but powerful Extended Horizon Model Algorithmic Control so that open-loop stable non-minimum phase systems can be effectively controlled in the time-varying environment. In the new, cascade structure of the ALS estimator, the covariance windup and blowup are totally eliminated. Moreover, the sacramental square-root update of the covariance matrix is no longer needed On the other hand, employing EHMAC facilitates robustness design so that controller parameters can be effectively auto-tuned.

Synthesis and Evaluation of New Ethyl 2,3- Dihydro-3-oxoisothiazolo[5,4-b]pyridine- 2-alkanoate Derivatives as Potential Herbicides

Several ethyl 2,3-dihydro-3-oxoisothiazolo[5,4-b]pyridine-2-alkanoate derivatives were synthesized as herbicides. Only 5-methyl derivatives inhibited both hypocotyl and root growth in the lettuce (Lactuca sativa L.) seedling test at 100 mg litre -1 . Only ethyl propionate and valerate derivatives showed significant inhibition at 0.1 mg litre -1 , whereas ethyl acetate or butyrate derivatives were inactive. Contrary to unoxidized derivatives, the inhibitory effect of 1-oxide and 1,1-dioxide derivatives was strongly dependent on concentration; ethyl 2,3-dihydro-5-methyl-3-oxoisothiazolo[5,4-b]pyridine-2-propionate 1,1-dioxide inhibited 100% of germination at 100 mg litre -1 and 45% of lettuce seedling growth at 0.1 mg litre -1 . Quantitative structure-inhibition of growth relationship analysis carried out by adaptive least-squares (ALS) method gave a good correlation with small and hydrophobic 5-substituents as well as with odd carbon-chain ethyl alkanoates in position 2. Active compounds did not show auxin-like activity from 0.1 to 100 mg litre -1 .

Higher order cumulants-based least squares for nonminimum-phase systems identification

A third-order cumulants based adaptive recursive least-squares (CRLS) algorithm for the identification of time-invariant nonminimum phase systems, as well as time-variant nonminimum phase systems, has been successfully developed. As higher order cumulants preserve both the magnitude and the phase information of received signals, they have been considered as powerful signal processing tools for nonminimum phase systems. In this paper, the development of the CRLS algorithm is described and examined. A cost function based on the third-order cumulant and the third-order cross cumulant is defined for the development of the CRLS system identification algorithm. The CRLS algorithm is then applied to different moving average (MA) and autoregressive moving average (ARMA) models. In the case of identifying the parameters of an MA model, a direct application of the CRLS algorithm is adequate. When dealing with an ARMA model, the poles and the zeros are estimated separately. In estimating the zeros of the ARMA model, the construction of a residual time-series sequence for the MA part is required. Simulation results indicate that the CRLS algorithm is capable of identifying nonminimum phase and time-varying systems. In addition, because of the third-order cumulant properties, the CRLS algorithm can suppress Gaussian noise and is capable of providing an unbiased estimate in a noisy environment.

Prediction of the rodent carcinogenicity of organic compounds from their chemical structures using the FALS method.

Fuzzy adaptive least-squares (FALS), a pattern recognition method recently developed in our laboratory for correlating structure with activity rating, was used to generate quantitative structure-activity relationship (QSAR) models on the carcinogenicity of organic compounds of several chemical classes. Using the predictive models obtained from the chemical class-based FALS QSAR approach, the rodent carcinogenicity or noncarcinogenicity of a group of organic chemicals currently being tested by the U.S. National Toxicology Program was estimated from their chemical structures.

On inverse factorization adaptive least-squares algorithms

This paper presents an adaptive exponentially weighted algorithm for least-squares (LS) system identification. The algorithm updates an inverse ‘square root’ factor of the input data correlation matrix, by applying numerically robust orthogonal Householder transformations. The scheme avoids, almost entirely, costly square roots and divisions (present in other numerically well-behaved adaptive LS schemes) and provides directly the estimates of the unknown system coefficients. Furthermore, it offers enhanced parallelism, which leads to efficient implementations. A square array architecture for implementing the new algorithm, which comprises simple operating blocks, is described. The numerically robust behavior of the algorithm is demonstrated through simulations. The algorithm is compared to the recently developed inverse factorization QR scheme (Alexander and Ghirnikar, 1993), in terms of computational complexity, parallel potential and numerical properties.

TEM Mapping Along the Longitudinal Valley in Eastern Taiwan and Their Tectonic Implications

The Transient Electromagnetic (TEM) method was used to map geoelectric structures beneath the Longitudinal Valley, eastern Taiwan. Coincident loop configurations with square transmitting loops 90 to 200m on each side were used for this survey. More than 300 soundings were conducted in this valley during 1991-1993, Field data were inverted by the adaptive nonlinear least-squares technique. The sounding depths were more than 200m beneath the Longitudinal Valley. A 3D resistivity image of the Longitudinal Valley was obtained. The resistivity patterns in the western part of the Longitudinal Valley are apparently higher and relatively more heterogeneous, in contrast with a resistivity which is on average lower and relatively less homogeneous in the eastern part of the Longitudinal Valley. This feature correlates well with the surface geology of the area. Moreover, the TEM soundings show that one predominant NNE-trending electrical transition in the Longitudinal Valley agrees well with the known suture trace on surface between the Philippine Sea and the Eurasian plate. On the basis of the resistivity pattern at different depths, the crust in this area is subject to a horizontal compressional stress oriented preferentially along the NWW-SEE direction. That the Longitudinal Valley is undergoing severe compression by the Philippine Sea plates can also be recognized. However, due to the limited depth of exploration, deeper investigations are needed to completely resolve the actual tectonic features.

QSAR(定量的構造活性相関)解析手法の開発とその応用

QSAR(定量的構造活性相関)解析手法の開発とその応用

Time-variant displacement structure and triangular arrays

The authors extend the concept of displacement structure to time-variant matrices and use it to efficiently and recursively propagate the Cholesky factor of such matrices. A natural implementation of the algorithm is via a modular triangular array of processing elements. When the algorithm is applied to solve the normal equations that arise in adaptive least-squares filtering, they get the so-called QR algorithm, with the extra bonus of a parallelizable procedure for determining the weight vector. It is shown that the general algorithm can also be implemented in time-variant lattice form; a specialization of this result yields a time-variant Schur algorithm. >