Religion & Spirituality

A Novel Fast Orthogonal Search Method for design of functional link networks and their use in system identification

Description
A Novel Fast Orthogonal Search Method for design of functional link networks and their use in system identification
Published
of 5
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  A Novel Fast Orthogonal Search Method for Design of FunctionalLink Networks and their Use in System Identification Hazem M. Abbas  Abstract —In this paper, a functional link neural net (FLN)capable of performing sparse nonlinear system identificationis proposed. Korenberg’s Fast Orthogonal Search (FOS) isadopted to detect the proper model and its associated parame-ters. The FOS algorithm is modified by first sorting all possiblenonlinear functional expansion of the input pattern accordingto their correlation with the system output. The sorted functionsare divided into equal size groups, pins, where functions withthe highest correlation with the output are assigned to thefirst pin. Lower correlation members go the following pinand so forth. During the identification process, members inlower pins are tried first. If a solution is not found, nextpins join the candidates pool until the identification processcompletes within prespecified accuracy. The modified GramSchmidt orthogonalization and Choleskey decomposition areapplied to create orthogonal functionals that can linearly fit theidentified system. The proposed architecture is tested on noise-free and noisy nonlinear systems and shown to find sparsemodels that can approximate the experimented systems withacceptable accuracy. I. I NTRODUCTION Identification of nonlinear dynamic systems is of consid-erable importance in many engineering applications such asecho cancellation [1], device modeling [2], nonlinear filterdesign [3] among others. Feedforward neural networks havebeen employed in system identification using a set of delayedinputs and outputs [4]. Multilayer perceptrons (MLP) trainedusing the backpropagation algorithm for system modelingwas presented in [5]. Other neural models also include radialbasis function nets [6] and orthonormal nets with Lagendrefunctions [7].Functional link networks (FLN) [8] replaces the hiddenlayer in MLP by providing nonlinear function expansion of the network input using functional links, polynomial basisfunctions, for example. The net output is composed of alinear sum of the basis functions. They have proved capableof approximating nonlinear mappings and were shown tosuccessfully model nonlinear systems [4].In system identification, the objective is to determinethe significant model terms and their associated values.Normally, most of the model terms are not contributing tothe system output resulting in a sparse model representation.When a FLN is trained to identify a nonlinear system,all candidate terms represented by the functional links aretreated equally. The value of all terms are estimated althoughonly a few terms are needed to be identified and thusresulting in a great waste of computing power. In addition, The author is with Mentor Graphics Egypt, 51 Beirut St., Heliopolis,Cairo 11571, Egypt (email: h.abbas@ieee.org). the estimated model parameters will become inaccurate sincea large number of insignificant terms are involved.Many methods have been introduced to address the modelselection problem such as evolutionary algorithms ([9], [10])and orthogonal least squares methods ([11], [12]). Evolu-tionary techniques rely on applying methods borrowed fromnature by applying genetic operators to find the correctmodel. Orthogonal techniques use the pool of candidateterms to calculate a new set of orthogonal terms that reducethe squared error between the system and the model.This paper presents a method to construct a minimum FLNnetwork with polynomial functional links by employing amodified FOS [11] that sorts the candidate terms accordingwith their correlation with the identified output so that termswith the highest correlation coefficients are used to formthe orthogonal space. Analysis and simulation results of theproposed method demonstrate the efficacy of the method infinding the optimal design of FLN in a time much shorterthan when the conventional FOS is used.The paper is organized as follows. The characterizationof the system identification problem is presented in sectionII. Section III presents a description of using FLN nets forsystem identification. Section IV reviews the FOS algorithmwhile section V introduces the sorted version of the FOS andits utilization to find the minimum polynomial expansions of the FLN. Analysis and experimental results of the proposednetworks are presented in section VI.II. T HE  N ONLINEAR  S YSTEM  I DENTIFICATION  P ROBLEM The nonlinear system identification problem is depictedin Fig. 1. The system has an input signal,  x ( n ) , and pro-duces an output,  y ( n ) ,  n  = 1 , ···  ,N,  where  N   is therecord length. The identification model should approximate  .                       SystemModel x ( n )  y ( n )ˆ y ( n ) + e ( n ) Fig. 1. Identification System the system output when excited with the same input by 2743 1-4244-0991-8/07/$25.00/©2007 IEEE  minimizing the error,  J   =  N i =1  y ( i ) −  ˆ y ( n )  2 , betweenthe observed output,  y ( n ) , and the modeled output,  ˆ y ( n ) ,within an acceptable accuracy. A model that is describedby nonlinear autoregressive with exogenous inputs can beexpressed as ˆ y ( n ) =  F   y ( n − 1) ,...y ( n − K  ); x ( n ) ,...,x ( n − L )  + e ( n ) where  F  ( . )  is a nonlinear function to be determined,  e ( n ) is  iid   model error, and  K   and  L  are maximum lags in theinput and output, respectively. When neural nets are used asthe model, the nonlinearity,  F  , is replaced by the output of the sigmoidal hidden layer(s) in MLP [4], Gaussian basisfunctions in RBF [13], Chebyshev basis functions [14], orhigher-degree polynomial in FLN [15].III. F UNCTIONAL  L INK  N ETWORKS  (FLN)FLN nets approximate a desired single output of asparse system using a small set of basis functions  {  p i ,i  =0 , ··· M  } ,  M    M ,  and a set of associated weights, { w i ,i  = 0 , ··· M  }  so that the output is expressed as ˆ y ( n ) = M   i =0 w i  p i ( n ) where M  is the total number of functional links that can becomposed by the network,  M   is the number of functionallinks that can reproduce the sparse system,  p 0  = 1  and allbasis functions are linearly independent. Figure 2 shows aFLN network approximating a single output. Multidimen-sional function approximation using FLN nets have beenintroduced in [16] and analyzed in more details in [15].                                                      u 1 u 2 u I   p 0 w 0  p 1  p M w 1 w M  + ˆ yy FunctionalExpansion Fig. 2. Functional Link Network Structure Assume that there is an  N   number of inputoutput patternpairs to be learned by the FLN, and the the input vector, U   ∈ R d , is composed of all possible  d  lags in the input andoutput, i.e.,  d  =  L  +  K  , while the resultant output,  ˆ y  is ascalar value. Each of the input pattern is passed through afunctional expansion block producing a corresponding M +1 -dimensional  ( M≥ d )  expanded vector. Considering all  N  patterns, inputoutput relationship may be expressed as ˆy  =  P w T  where  P  = [ p 0  p 1 ··· p M +1 ]  is a  N  × ( M +1)  dimensionalmatrix,  p i  is an  N  -dimensional vector representing the i th basis function, and  ˆy  is an  N  -dimensional estimatedoutput. In order to find the weight vector,  w , a number  N  of simultaneous equations need to be solved. If the basisfunctions were chosen to be formed as nonlinear polynomialsof the lagged input and output samples, the weights canbe calculated using the least square error method based onthe vector of the output signals,  y , and the regressor datarepresented by  P , i.e., w  = ( P T  P ) − 1 P T  y Since the least square methods consider each polynomialterm equally important and estimate all associated weightseven when only a small subset of those polynomials needto be identified (which is the case with in sparse sys-tems). Hence, this amounts to a huge waste of computationresources. Additionally, estimating a large number of in-significant terms will introduce inaccuracies in the estimatedterms. Orthogonal least squares (OLS) search techniques[12] alleviate this problem by applying forward stepwisemodel selection techniques to find the significant terms. Allterms are studied to determine the amount of contribution of each term in modeling the desired system. Several methodshave been proposed to improve the OLS techniques such asKorenberg’s Fast Orthogonal Search (FOS) [11] and [17]. Inthis work, a variation in the FOS is proposed to make it morecomputationally efficient in finding the exact model.IV. F AST  O RTHOGONAL  S EARCH  (FOS)[11]The FLN output can be expressed as a time series ˆ y ( n ) = M   i =0 w i  p i ( n ) + e ( n )  (1)where  p 0 ( n ) = 1  and for  m ≥ 1 ,  p m ( n ) =  x ( n − l 1 ) x ( n − l j ) ··· y ( n  −  k 1 ) y ( n  − k i ) , and  0  ≤  l v  ≤  L ,  0  ≤  k v  ≤ K  . Using the orthogonal search method, the model in (1) isexpressed as ˆ y ( n ) = M   i =0 g i  s i ( n ) + e ( n )  (2)The new orthogonal basis,  s i ( n ) , are constructed from the  p i ( n )  using the modified Gram-Schmidt procedure so thatthey are orthogonal over the observation period of the output.The parameters,  g i , are selected to minimize the mean–squared error over the interval, i.e., e 2 ( n ) =  y ( n ) − M   i =0 g i  s i ( n )  2 =  y 2 ( n ) − M   i =0 g 2 i  s 2 i ( n ) (3)The overbar denotes the time average. It can be easily shownthat the addition of a new term,  w r  p r ( n ) , will reduce theerror by the amount  Q ( r ) =  g 2 r  s 2 r ( n )  where g r  =  y ( n ) s r ( n ) s 2 r ( n ) .  (4) 2744  In order to expand the model, it is required to calculatethe quantity  Q  for all candidates and choose the one forwhich  Q  is the greatest. Note also that the construction of the orthogonal functions,  s i ( n ) , is computationally intensiveas it should be done for each candidate term [18]. The FastOrthogonal search avoids this problem. Using Gram-Schmidtorthogonalization, the functions  s i ( n )  are created as follow: s 0 ( n ) = 1 s r ( n ) =  p r ( n ) − r − 1  i =0 α ri  s r ( n ) ,r  = 1 , ···  ,M   (5)where α ri  =  p r ( n ) s i ( n ) s 2 i ( n ) , i  = 0 , ···  ,r − 1 Define  D ( m,r ) =  s m ( n ) s r ( n )  and using (5), D ( m,r ) =  p m ( n )  p r ( n ) − r − 1  i =0 α ri  D ( m,i ) ,  (6) D ( m, 0) =  p m ( n ) ,r  = 1 , ···  ,M  ; m  = 1 , ···  ,M   (7)Similarly, s 2 m ( n ) =  p 2 m ( n ) − m − 1  r =0 α 2 mr  s 2 r ( n ) ,m  = 1 , ···  ,M   (8)and E  ( m ) =  s 2 m ( n ) , E  (0) = 1 ,m  = 0 , ···  ,M   (9)This results in α mr  =  D ( m,r ) E  ( r )  ,r  = 0 , ···  ,m − 1; m  = 1 , ···  ,M   (10)This provides two recursive equations D ( m,r ) =  p m ( n )  p r ( n ) − r − 1  i =0 D ( r,i )  D ( m,i ) E  ( i )  ,  (11) r  = 1 , ···  ,m − 1; m  = 2 , ···  ,M E  ( m ) =  p 2 m ( n ) − m − 1  r =0 D 2 ( m,r ) E  ( r )  ,  (12) m  = 1 , ···  ,M  By repeating a similar procedure with the output,  y , one willobtain another recursive equation, C  ( m ) =  y ( n )  p m ( n ) − m − 1  r =0 α mr  C  ( r ) ,m  = 1 ···  ,M   (13)and  C  (0) = 1 . Using (13) and (12), g m  =  C  ( m ) E  ( m ) ,m  = 0 ···  ,M   (14) Q ( m ) =  g 2 m  E  ( m )  (15) e 2 ( n ) =  y 2 ( n ) − M   m =0 g 2 m  E  ( m )  (16)The  M   selected basis functions are those that produce thelargest  Q ( m )  in (15). Finally, the weight values associatedwith the selected functional links,  w i , can be calculated usingthe following [19]: w m  = M   i = m g i v i  (17) v m  = 1 , v i  = − i − 1  r = m α ir v r  (18)This completes the identification of the FLN basis functionsneeded to represent the sparse system and their associatedweight values.The speed up offered by this algorithm is achieved byexploiting the lagged nature of the difference equation in the  p i ( n )  terms that makes it possible to accelerate the calcu-lations of the different time averages in the FOS algorithm.This is done by relating the time averages to input and outputmeans and correlations and then by making small correctionsfor the finite record length. More details can be found in[20]. The method has shown to save a lot of computationtime and memory storage. Also, the recursive equations in(12) and (13) requires calculations to be performed for thecurrent candidate while values for previous candidates arereused.V. S ORTED  F AST  O RTHOGONAL  S EARCH  (SFOS)The FOS can be enhanced further through different vari-ations. This can be done by arranging the basis functionsterms in such a way that most  probable  terms are selectedfirst. A method was proposed in [21] where the terms aregrouped into disjoints subsets that are searched sequentially.The set with linear  x  terms are searched first, followed by the y  terms, and then by the nonlinear  xx  terms, the  yy  terms,and finally by the  xy  terms. A similar approach was proposedin [22] in their genetic evolution of a FLN by favoringsimple models first. More complex nonlinear individuals aresearched when the simpler linear terms become unable to fitthe required mapping.This work presents another modification to the FOS al-gorithm. The main idea of the proposed modification isto exploit the fact that basis functions with the highestcorrelation with output are more likely to be principal systemterms [10]. The algorithm starts by sorting all candidatebasis functions,  M , in descending order according to theircorrelation with the output and grouping them into a numberof pins, V   . Each pin is assigned an equal number of candidatefunctions, R  =  M V    , where candidates with highest correlationwith the output go the first pin and lower correlation termsare assigned to the following one and so forth. Then theconventional FOS is applied to operate on the candidates inthe first pin. Obviously, the majority of the candidates neededto fully represent the system will be picked there and a greatreduction in the representation error is expected. If a solutionwithin acceptable accuracy cannot be attained after testing allfirst pin candidates, the members of the second pin are added 2745  to the candidate pool and the FOS algorithm continues untila final acceptable solution is reached.Since the term with the largest error reduction contribution, Q ( r ) , should be selected first, moving the terms that arehighly correlated with the output higher up in the candidatelist, will result in a much faster convergence to the minimumbasis function architecture. To demonstrate the validity of thisconjecture, let us examine the procedure of selecting the firstterm. To calculate  Q (1) , we start with the values of the firstconstant basis function, m  = 0 , E  (0) = 1 , C  (0) =  y, D (1 , 0) =  p 1 ( n ) The reduction in error by the chosen term is, Q (1) =  g 21  s 21 ( n ) =  g 21  E  (1) =  C  2 (1) E  (1) Using (13) and (12), C  (1) =  y ( n )  p 1 ( n ) − α 10 C  (0) =  y ( n )  p 1 ( n ) − y ( n )  p 1 ( n ) E  (1) =  p 21 ( n ) − α 10 D (1 , 0) =  p 21 ( n ) − [  p 1 ( n )] 2 This gives, Q (1) =  y ( n )  p 1 ( n ) − y ( n )  p 1 ( n )  p 21 ( n ) − [  p 1 ( n )] 2 =  y ( n )  p 1 ( n ) − y ( n )  p 1 ( n ) σ 2  p 1 (19)where  σ 2  p 1 is the variance of the term,  p 1 ( n ) . The correlationcoefficient,  ρ ( y,p 1 ) , between the output and the tested termis defined as ρ 2 ( y,p 1 ) = [( y ( n ) − y ( n ))(  p 1 ( n ) −  p 1 ( n ))] 2 σ 2  p 1 σ 2 y It is straightforward to show that σ 2  p 1 Q (1) =  ρ 2 ( y,p 1 )  σ 2 y  σ 2  p 1 and hence Q (1) ∝ ρ 2 ( y,p 1 )  (20)Therefore, the candidate with the highest  Q (1) , is the termwhich has the maximum  ρ 2 ( y,p 1 )  with the observed output.This finding justifies the proposed ordering of the candidates.By removing the output contribution of this candidate fromthe srcinal output and repeating the process with the re-maining candidates, the above analysis will always select thecandidate with the highest correlation with the new output.It should be noted that such ordering does not guaranteethat all significant candidates will be placed high on the list.However, the CPU time needed by the FOS algorithm is theupper bound of the proposed SFOS. As it will be shown inthe experiments, the SFOS will be shown to perform muchfaster than FOS. TABLE IE XAMPLE  1: O RDERING OF  B ASIS  F UNCTIONS BEFORE AND AFTER S ORTING No. Candidate location loc. after sorting1 20 32 118 73 282 784 357 125 403 46 411 137 715 18 822 769 831 510 1495 2 VI. S IMULATION  R ESULTS AND  A NALYSIS The proposed FLN-SFOS identification algorithm has beentested using a set of experiments. We will describe twoexamples to demonstrate the performance of the algorithmin identifying a sparse 3- rd   order system. The locationsof the principal basis functions are chosen randomly. Thecorresponding weight values of the chosen functions aregenerated from a uniform distribution bounded by the upperand lower values of each kernel  ([ − 2 . 5 , 2 . 5]) . The firstexample demonstrates the ability of the algorithm in correctlyidentifying the system driven by a white Gaussian input andnoise-free measurement. The second example describes thealgorithm performance when there is a white Gaussian mea-surement noise. In the experiments, the input sequence, x ( n ) ,is drawn from a zero-mean unit-variance white Gaussiandistribution and the output is observed for a record lengthof   500+( K  + L ) . The first  K   + L  samples are discarded toeliminate any error that might result in calculating the timeaverages.  A. Example 1: A third-order system with no measurement noise The first example is of a 3–rd order system with timedelays equal to  ( L  = 21; K   = 0) .  This amounts to a totalnumber of basis functions,  M  = 2024 . Only 10 candidates ( M   = 10   M )  have been used to generate the output.A pin size of   R  = 10 , is used which results in having21 pins to try. Table I shows the ordering of the selectedbasis functions before and after the correlation-based sortingprocess. It is obvious that one basis function is linear andthe rest are combinations of nonlinear terms in 2– nd   and3– rd   order terms of   x . The sorting process resulted in allsignificant terms to be grouped within the first pin (highestcandidate is in the 78th position). If the conventional FOS isapplied on this data, a number of 1495 terms has to be testedin order to completely reconstruct the srcinal system. Withthe SFOS, the process is terminated after the 78th term istested and even before trying all candidates in the pin. Theorder of selection of the basis functions by the SFOS is 1,2, 3, 7, 5, 4, 12, 13, 78, and 76. Obviously, the terms withthe highest  ρ 2 ( y,p i )  were selected first. When the FOS isapplied on the system, it required 2.2969 sec of computingtime. The proposed SFOS-FLN algorithm only took 0.093 2746  sec. of computing time, a significant speed–up factor of 24.Expectedly, both algorithms reached the same representationerror.  B. Example 2: A third-order system with Gaussian noiseadded  The previous experiment was repeated after adding a whiteGaussian noise to the output with a SNR equal to 10 dB.As in the first example, the input used is a 500 data pointsdrawn from a white Gaussian noise with zero-mean and unitvariance. The measurement noise is independent from theinput signal. Due to noise, it is expected that extra basisfunctions are added in the final solution as an attempt to fitthe noise. The FOS managed to identify the srcinal basisfunctions. However, the corresponding weights,  w i  weredifferent from the srcinal one. Moreover, nine extra basisfunctions were added to the model in order to achieve afinal mse error equal to 1.528. After candidate sorting, onceagain all candidates were placed in the first pin. Naturally,the exact locations within the pin differed from the one inthe noise–free case due to the new correlation values causedby the noisy output. The SFOS managed to identify the 10srcinal candidates with exceptional accuracy in the weightvalues and only extracted just one extra false candidate witha very small weight value. However, the final mse error was1.7, an 11% increase than the mse obtained by the FOS. TheCPU time needed for the noisy case was 0.1094 sec. for theSFOS and 2.2969 sec. for the FOS, a speed–up of about 22.It is worth mentioning that the speed–up factor is dependenton the number of candidate functionals, M , and the pin size, V   .The SFOS-FLN need also to be tested on classificationproblems to test the capability of the algorithm in producingnonlinear discriminant analysis functions based on the higherpolynomials offered by the FLN.VII. C ONCLUSIONS In this paper, an approach, SFOS, of constructing FLNfor sparse polynomial systems identification is presented.The proposed algorithm exploits the fact that FLN basisfunctions with the highest correlation with output are moreprobable to be principal system terms. The algorithm sorts allavailable candidates in descending order according to theircorrelation with the output and assigns them to different pinsof fixed size. It has been shown that such an ordering willguarantee that principal terms will get selected first. The FOSalgorithm is applied to pick up the correct candidates andcalculates their associated weight values using orthogonalsearch and Cholskey decomposition. The SFOS algorithmexamines first the basis function candidates in the firstpin. If the selected functions become unable to produce anacceptable solution, subsequent candidates in the followingpins are added to the pool. The process ends when a solutionis found and the identified system is successfully reproduced.In the absence of any measurement noise, the algorithmhas produced exact results when applied to sparse secondand third order systems. For noisy outputs, the algorithmmanaged to detect the correct terms with a small error inthe kernel values. In either case, the SFOS has shown toprovide a considerable speed–up factor when compared withthe FOS.R EFERENCES[1] O. Agazzi and D. G. Messerschmitt, “Nonlinear echo cancellationof data signals,”  IEEE Transcation on Communication , vol. COMM,no. 30, pp. 2421–2433, 1982.[2] G. Stegmayer, “olterra series and neural networks to model anelectronic device nonlinear behavior,” in  Proceedings of the IEEE  International Joint Conference on Neural Networks (IJCNN) , 2004,pp. 2907–2910.[3] J. D. Taft, “Quadratic linear filters for signal detection,”  IEEE Tran-scation on Signal Processing , vol. SP, no. 39, pp. 2557–2559, 1991.[4] S. Chen, S. Billings, and P. Grant, “Nonlinear system identificationusing neural networks,”  International Journal of Control , vol. 51, no. 6,pp. 1191–1214, 1990.[5] K. Narendra and K. Parthasarathy, “Identification and control of dynamical systems using neural networks,”  IEEE Transactions on Neural Networks , vol. 1, pp. 4–27, 1990.[6] S. Elanyar and Y. Shin, “Radial basis fnction neural network for ap-proximation and estimation of nonlinear stochastic dynamic systems,”  IEEE Transactions on Neural Networks , vol. 5, pp. 594–603, 1994.[7] S. Yang and C. Tseng, “An orthonormal neural network for functionapproximation,”  IEEE Transactions on Systems, Man and Cybernetics,Part B , vol. 26, pp. 779–785, 1996.[8] Y.-H. Pao, S. Phillips, and D. Sobajic, “Neural-net computing andintelligent control systems,”  International Journal of Control , vol. 56,no. 2, pp. 263–289, 1992.[9] L. Yao, “Genetic algorithm based identification of nonlinear systemsby sparse volterra filters,”  IEEE Transactions on Signal Processing ,vol. 47, no. 12, pp. 3433–3435, 1999.[10] H. Abbas and M. Bayoumi, “Volterra-system identification usingadaptive real-coded genetic algorithm,”  IEEE Transactions on Systems, Man and Cybernetics, Part A , vol. 36, no. 4, pp. 671– 684, 2006.[11] M. J. Korenberg, “A robust orthogonal algorithm for system identifi-cation and time–series analysis,”  Biological Cybernetics , vol. 60, pp.267–276, 1989.[12] S. Chen, C. Cowan, and P. Grant, “Orthogonal least squares learningfor radial basis function networks,”  IEEE Transactions on Neural Networks , vol. 2, no. 2, pp. 302–309, 1991.[13] S. Chen, S. Ching, and K. Alkadhimi, “Regularized orthogonal leastsquares algorithm for constructing radial basis function networks,”  International Journal of Control , vol. 64, no. 5, pp. 829–837, 1996.[14] S. Purwara, I. Karb, and A. Jha, “On-line system identification of complex systems using chebyshev neural networks,”  Applied Soft Computing , vol. 7, no. 1, pp. 364–372, 2007.[15] J. Patra, R. Pal, B. Chatterji, and G. Panda, “Identification of nonlineardynamic systems using functional link artificial neural networks,”  IEEE Transactions on Systems, Man and Cybernetics, Part B , vol. 29,pp. 254–262, 1999.[16] T. Yamada and T. Yabuta, “Dynamic system identification using neuralnetworks,”  IEEE Transactions on Systems, Man and Cybernetics ,vol. 23, pp. 204 – 211, 1993.[17] E. Chng, S. Chen, and B. Mulgrew, “Efficient computational schemesfor the orthogonal least squares algorithm,”  IEEE Transactions onSignal Processing , vol. 43, pp. 373 – 376, 1995.[18] M. J. Korenberg and L. D. Paarmann, “Orthogonal approaches to time–series analysis and system identification,”  IEEE Signal Processing Magazine , pp. 29–43, July 1991.[19] M. J. Korenberg, S. Bruder, and P. McIlroy, “Exact orthogonal kernelestimation from finite data records: Extending wiener’s identificationof nonlinear systems,”  Annals of Biomedical Engineering , vol. 16, pp.201–214, 1988.[20] M. J. Korenberg, “Identifying nonlinear difference equation andfunctional expansion representations: The fast orthogonal algorithm,”  Annals of Biomedical Engineering , vol. 16, pp. 123–142, 1988.[21] P. McIlroy, “Applications of nonlinear system identification,” Master’sthesis, Queen’s University at Kingston, Ontario, Canada, 1986.[22] A. Sierra, J. A. Macias, and F. Corbacho, “Evolution of functional link networks,”  IEEE Transactions on Evolutionary Computation , vol. 5,no. 1, pp. 54–65, 2001. 2747
Search
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks