Alphanumeric Character Recognition

Size: px
Start display at page:

Download "Alphanumeric Character Recognition"

Transcription

1 Amerian Journa of Signa Proessing. 0; (): DOI: /j.ajsp Training Tangent Simiarities with N-SVM for Aphanumeri Charater Reognition Hassiba Nemmour *, Youef Chibani Signa Proessing Laboratory, Fauty of Eetroni, University of Houari Boumediene, Agiers, Ageria Abstrat This paper proposes a fast and robust system for handwritten aphanumeri harater reognition. Speifiay, a neura SVM (N-SVM) ombination is adopted for the assifiation stage in order to aeerate the running time of SVM assifiers. In addition, we investigate the use of tangent simiarities to dea with data variabiity. Experimenta anaysis is onduted on a database obtained by ombining the we known USPS database with C-Cube upperase etters where the N-SVM ombination is evauated in omparison with the One-Against-A impementation. The resuts indiate that the N-SVM system gives the best performane in terms of training time and error rate. Keywords Handwritten Aphanumeri Charaters, Svms, Tangent Vetors. Introdution In various appiations of doument anaysis, aphanumeri harater reognition onstitutes a very important step. For instane, in bank hek proessing this task is required for date reognition whie in automati mai sorting it is used to reognize addresses. However, the fat that haraters are written in various manners by different sripts and different toos onduts to high simiarity between some upperase etters and digits suh as Z and or O and 0. Due to this ambiguity, assifiation systems fai ommony in disriminating upperase etters from digits. Besides, in handwriting reognition robust assifiers shoud be used to ahieve a satisfatory performane sine onventiona ones provide systematiay insuffiient reognition rates[]. In the past reent years, attentions were foused on earning mahines suh as neura networks and hidden Markov modes[]. Currenty, Support Vetor Mahines or SVMs are the best andidate for soving a handwriting reognition tasks with medium number of asses[3]. Speifiay, SVMs were used for handwritten digit reognition in many researh works[4,5]. In this work, we investigate their use for soving an aphanumeri harater reognition whih aims to disriminate upperase Latin etters from digits. However, it is obviousy known that suh appiation handes ommony arge sae databases whereas the SVM training time is quadrati to the number of data. For this reason, we investigate aso, the appiabiity of N-SVM ombination for aphanumeri assifiation. N-SVM is a neura-svms ombination * Corresponding author: hnemmour@yos.om(hassiba Nemmour) Pubished onine at Copyright 0 Sientifi & Aademi Pubishing. A Rights Reserved whih was introdued for handwritten digit reognition in order to redue the runtime and improve auray[6]. Furthermore, due to the arge variabiity of aphanumeri haraters the use of feature extration shemes is a prerequisite to reah suffiient reognition auraies. Speifiay, feature extration an aeviate some imitations reated to sript variations as we as to some segmentation errors. Many handwriting reognition researh works have shown that earier desriptors whih were extensivey used for pattern reognition suh as geometri moments and generi Fourier desriptors annot dea with haraters ambiguity[7,8]. Thereby, more effiient desriptors were deveoped to aow invariane with respet to some variations of data. Reenty, some features based on urvature or ontour information were introdued for haraters reognition. Among them, we note the ridgeet transform whih showed high disrimination abiity for printed Chinese haraters[9]; urvature features for handwritten digit reognition[0] and waveet pakets for handwritten Arabi word reognition[]. In fat, a good desriptor shoud be invariant with respet to some affine transformations suh as sma deformations, transations and rotations whie being variabe from a ass to other. In this framework, tangent vetors whih are based on invariane earning onstitute one of the best desriptors for handwritten digit reognition[,3]. Tangent vetors are omputed by performing a set of affine transformations to eah pattern. Then, assifiers are trained on the generated tangent sampes to produe invariant deision funtions. Sine the resuts for handwritten digits were very promising, presenty, we try to inorporate tangent vetors into the proposed aphanumeri harater reognition system. Speifiay, tangent simiarities based on ass-speifi tangent vetors are used for both desribing variabiity and reduing the size of data[4]. The rest of this paper is arranged as foows. Setion

2 Amerian Journa of Signa Proessing. 0; (): gives a brief review of SVM assifiers. Setion 3 desribes the N-SVM arhiteture whie setion 4 presents the tangent vetor based simiarities. The ast setions summarize the experimenta resuts and give the main onusions of the paper.. Support Vetor Mahines Support Vetor Mahines (SVMs) were designed to onstrut binary assifiers from a set of abeed training sampes defined by: ( x i, yi ) R N { ±}, where i =, ( is the number of training data). SVMs seek the inear separating hyperpane with the argest margin by soving the foowing optimization probem[3,5]: T Minimize w w () Subjet to yi ( xi w + b) i () T denotes the transpose, b is a bias whie w is the norma to the hyperpane. When inequaities in (3) do not hod for some data, the SVM is non-ineary separabe. Then, the margin of separation is said to be soft and non-separabe data are handed by introduing a set of nonnegative sak variabes { ξ i } into the deision surfae[3]. Then, the goa is to find a hyperpane whih minimizes misassifiations whie maximizing the margin of separation suh that: T Minimize ww+ C ξi (3) i= Subjet to yi ( xi w + b) ξi (4) C is a user-defined parameter that ontros the tradeoff between the mahine ompexity and the number of non-separabe data[4]. Commony, a dua Lagrangian formuation of the probem in whih data appear in the form of dot produts, is used: Maximize LD = αi αα i jyyx i j i xj (5) Subjet to i i= i, j α iyi = 0 (6) where α i are Lagrange mutipiers. The dua probem is usefu when data are not ineary separabe in input spae. In suh a ase, they are mapped into a feature spae via a kerne funtion suh that: K( xi, x) = φ( xi ), φ( x). The dua probem beomes: LD = αi αα i jyyk i j ( xi, xj) (7) i Thereby, the deision funtion is expressed in terms of kerne expansion as: Sv i= i, j f( x) = αiyk i ( xi, x) + b (8) Sv is the number of support vetors whih are training data for whih 0 < α j C. The optima hyperpane orresponds to ( x) = 0 f whie test data are assified aording to: ( x) ( x) positive ass if f > 0 x (9) negative ass if f < 0 A mathematia funtions whih satisfy Merer s onditions are eigibe SVM-kernes. The simiarity between data is expressed either in the form of a dot produt or distane measure. The most ommony used kernes in pattern reognition are isted in Tabe (). Noyau Poynomia: Po Sigmoïde: Sig RBF Distane Négative: ND pp, ss 0, ss, σσ, γγ: are user defined. Tabe. Some SVM kernes KK xx ii, xx jj xx ii xx jj + pp tttttth ss 0 xx ii xx jj + ss eeeeee xx ii xx jj σσ xx ii xx jj γγ Furthermore, various approahes were proposed to extend SVMs for muti-ass assifiation probems[6]. The One-Against-A (OAA) is the eariest and the most ommony used impementation. For a C -ass probem, It performs M SVMs eah of whih is designed to separate a ass from a the others. The th SVM is trained with a of the exampes in the th ass with positive abes, and a other exampes with negative abes whih eads to a omputationa time approximatey about C. Then, data are assigned to the positive ass with the highest output as: C arg max f x (0) = ( ) Unfortunatey, the time required for training the OAA-based SVM is a imitation for arge sae databases. Rea that the training of a SVM is quadrati to the number of data[7]. This means that for OAA this time is mutipied by the number of asses. Thereby, for aphanumeri assifiation one must dea with this time ompexity. Presenty, we adopt a Neura-SVM (N-SVM) ombination whih aims to aeerate the runtime of SVMs. Sine this ombination was introdued for handwritten digit reognition we try to extend its use for aphanumeri harater reognition. The N-SVM is briefy presented in the foowing setion. 3. N-SVM Combination The goa of N-SVM ombination onsists of reduing the runtime required by standard SVMs. This arhiteture was proposed for handwritten digit reognition where 5 dihotomies are randomy seeted from the 45 possibe dihotomies between numera asses. Then, outputs of the SVMs trained over the seeted dihotomies are handed by a neura network to produe an automati deision about asses. The resuts obtained for USPS database highighted the vaidity of this approah to redue the runtime and improve the reognition auray[6]. Thus, for aphanumeri

3 36 Hassiba Nemmour et a.: Training Tangent Simiarities with N-SVM for Aphanumeri Charater Reognition harater reognition, whih is a probem of 36 asses (0 digits and 6 upperase etters), we deveop 8 SVMs designed to independenty separate pairs of asses. In summary, the N-SVM ombination is omposed of the foowing steps: From the C (C-)/ possibe dihotomies (where C=36): a) Seet randomy a dihotomy and assoiate its asses Ci and C j to a SVM. b) Deete asses C i and C j from the set of asses. ) Return to ) if the number of seeted pairs is ess than C/. d) Train SVM assifiers over the seeted pairs of asses. e) Train a feed-forward neura network over SVM outputs. The neura network is MutiLayer Pereptron (MLP) with a singe hidden ayer. The input ayer ontains 8 nodes, whih reeive SVM outputs whereas eah node in the output ayer orresponds to a numera ass from 0 to Z. On the other hand, the number of hidden nodes is fixed experimentay. The MLP is trained by the standard bak propagation agorithm inuding the momentum fator[8]. In addition, N-SVM system reeives feature vetors whih are onstituted by tangent simiarities with respet to a asses. The omputation of suh simiarities is detaied in the foowing setion. 4. Tangent Vetor Based Simiarities for SVMs A priori knowedge was initiay introdued to SVMs through the VSV approah whih generates virtua sampes from support vetors to enfore invariane around the deision boundary[9]. Indeed, invariane earning is based on the idea of earning sma or oa transformations, whih eave the ass of data unhangeabe. To define these transformations, et ~ x( β ) denotes a transformation of a pattern x that depends on a set of parameters L β = { β,, β L} R. The inear approximation of ~ x ( β ) using a Tayor expansion around β = 0 an be written as[4]: L L ( ) = x + β + ο( ) = v ~ x β β () = v are Tangent Vetors (TV) orresponding to partia derivatives of the transformation ~ x with respet to β = ( =, L) so that : ~ x( β ) v = β = 0 () β Sine terms of seond order and higher in () are ommony negeted, ~ x ( β ) = x for β = 0 whie for sma vaues the transformation does not hange the ass membership of x. Hene, the inear approximation desribes transformations suh as transation, rotation and axis deformations by one prototype and its orresponding tangent vetor. The first use of the TV was introdued through the so-aed Tangent distane with the K-NN assifier[]. This distane gave very good preisions for handwritten digit reognition but its time ompexity was very important. Aways for handwritten digit reognition, TV were used with SVM assifiers by introduing the Tangent distane into distane-based kernes suh as RBF and negative distane[0]. However, the runtime of SVMs was signifianty extended. More reenty, a simiarity and dissimiarity measures were introdued in order to aow using tangent vetors with a SVM kernes[]. These measures are based on ass-speifi tangent vetors whose auation is faster ompared to the onventiona sheme. Then, the time ompexity is redued but the runtime is sti extended. In the present work, we aim to introdue the TV onept for aphanumeri harater reognition without extending the runtime of SVMs and N-SVM. Speifiay, we empoy tangent simiarities whih are based on ass-speifi TV, as data features. The ass-speifi TV are omputed as foows[4]. For a set of asses { =, C} with training data x n { n =, N C }, tangent vetors maximizing the dissimiarity between asses are hosen suh that are the eigenvetors with argest eigenvaues { Σ } v of the matrix : Σ (3) S ( Σ ) T Σ : Covariane matrix omputed from the fu dataset. S : Cass dependent satter matrix given by: N S = x n= ( )( ) T x µ n µ (4) µ : mean of the ass. It is straightforward to not that the number of TV shoud be obviousy fixed by the user. In addition, eah tangent vetor refers to a transformation or speifi variabiity knowedge whose number shoud be a priori hosen by the user. So, one estimated tangent vetors are inorporated into the ovariane matrix speifi to their asses suh that[4]: ~ Σ = Σ Σ Σ L T v v (5) + = γ γ : user defined parameter. Σ ~ : TV-based ovariane matrix for the ass. Furthermore, the tangent simiarity orresponds to the Tangent Vetor-based Mahaanobis (TVM) distane of a pattern x with respet to a ass beomes as: n T ~ ( x ) = ( x µ ) Σ ( x ) TVM µ (6) In[4], these simiarities were used to improve the ikeihood funtion of Bayesian assifiers. For our part, we investigate their use as data features for training SVMs. The idea onsists of substituting eah pattern by a feature vetor P x = TVM x, TVM x C onstituted by its TVM with ( ) { ( ) ( )} respet to a asses. Then, to evauate kerne funtions the distane x P x i P x whie the x i is repaed by ( ) ( )

4 Amerian Journa of Signa Proessing. 0; (): dot produt is substituted by ( x ) P( x) P i. Note that benefits behind the use of suh feature vetors are not ony the impiit inorporation of invariane knowedge into SVMs but aso the redution of the data size. For aphanumeri harater reognition the data size dereases from the harater image size to a vetor of 36 omponents (whih is the number of asses). This makes the kerne evauation as we as the training stage faster. 5. Experimenta Resuts Experiments are onduted on a dataset obtained by ombining the we-known USPS handwritten digits with a set of ursive upperase etters extrated from the C-Cube (Cursive Charater Chaenge) database. The USPS is omposed of 79 training data and 007 test data distributed on the 0 digits. These images are normaized to a size of 6 6 pixes yieding 56 dimensiona datum vetor. For this reason, ursive upperase etters of C-Cube database whih ome with varying sizes were normaized to the same size. This database ontains 5793 ursive haraters manuay extrated from ursive handwritten words, inuding both upper and ower ase versions of eah etter. A set of 608 upperase etters were extrated and normaized to sae with the USPS data. The resuting database inudes 734 training sampes and 358 test sampes. Figure shows some exampes whih highight the simiarity between upperase etters and digits S O G H Figure. Training sampes from the USPS-C-Cube database. We first performed a series of OAA and N-SVM runs to test out a possibe onfigurations. Based on the orresponding error rates the best parameter vaues were seeted as foows. The reguarization parameter is fixed at 0 whie kerne parameters are as: the sigma is 5 for RBF, d= for poynomia kerne, γγ = for negative distane whie sigmoid parameters are tuned to: ss 0= 0.009, ss =.. The neura network of N-SVM is a mutiayer pereptron with 90 hidden nodes, the step size equas 0.08, the momentum is about and the network is trained for 5000 iterations eah of whih orresponds to one proessing of a training data through the network. In order to assess the behavior of tangent simiarities with kernes based on a dot produt and distane measure, we evauated the Error Rate (ER) of both OAA and N-SVM by using kernes isted in Tabe (). The resuts in terms of error rate are summarized in Tabe. In this test, the number of tangent vetors was arbitrariy fixed at 0. Tabe. Error rates (%) provided by the training of tangent simiarities using 0 tangent vetors RBF ND Po Sig N-SVM 8,5 5, 7,4 30, OAA 3,37 5,9 9,78 9,7 We an note that N-SVM outperforms the OAA with three kernes whie it is sighty ess aurate when using the sigmoid kerne. In addition, with both approahes, distane-based kernes dea better than dot produt-based kernes where the RBF gives the best resuts. This kerne aows a gain whih equas at east 7%. In the seond test, we investigated the behavior of OAA and N-SVM with respet to variations of tangent vetors number. This evauation was arried out by using the RBF kerne with the preedent parametri seetion. So, for both approahes, variations of error rate for different hoies of the TV number are potted in Figure. Roughy speaking, whatever the number of TV, OAA and N-SVM behave simiary but with different error rates. More preisey, the N-SVM ombination gives a smaer error rate with a differene of 0.5%. We note aso, the first TV improve the error rate whih dereases with a fator of more than 4% with N-SVM and 6% with OAA. However, arger numbers of TV have a ompounding effet sine the error rate grows to more than 3% with both approahes. This outome an be expained by the fat that TV are obtained by eigenvetor deomposition. So, simiary to the prinipe omponent anaysis the ast eigenvetors are noisy and annot bring any disriminating knowedge about data. Error rate (%) Error rate variations aording to the number of tangent ve- Figure. tors N-SVM OAA Number of Tangent Vetors Finay, the best error rates are obtained for 30 TV. The resuts orresponding to this hoie are reported in Tabe 3. In addition to error rates, this tabe reports the number of support vetors (#SV) per binary node (SVM), the Training Time (TT) expressed in hours as we as the reognition speed whih is the Number of Reognized Charaters (NRC) per seond. As an be seen, N-SVM outperforms the OAA by 0.8% in error rate whie being muh faster. In fat, N-SVM training is 5 times faster beause it empoys 8 SVMs trained on different dihotomies whie the neura network takes about 5 minutes. On the ontrary, eah SVM

5 38 Hassiba Nemmour et a.: Training Tangent Simiarities with N-SVM for Aphanumeri Charater Reognition in the OAA is trained over the fu dataset whih yieds a arger runtime. In onsequent, N-SVM requires a smaer number of SV per SVM node the reason for whih it has a higher reognition speed. Hene, roughy speaking N-SVM provides a signifiant aeeration of the runtime whie being sighty more aurate. Tabe 3. Performane evauation of N-SVM and OAA using 30 TV OAA N-SVM Error rate (%) #SV TT (Hours) NRC (seond) Furthermore, Figure 3 exhibits the error rates ahieved in eah ass of interest by OAA and N-SVM. We remark that error rates respetive to numera asses are ommony smaer than those ahieved for severa upperase etters whih an reah 00%. This is the ase of etters I, J and Z whih are ompetey onfused with other asses when using the OAA approah. This outome is reated to the number of training sampes whih is smaer (ess than 0 training sampes) ompared to those of the other asses. In fat, in the onsidered dataset etter asses have smaer training sets ompared to numera asses. In addition, the normaization of etter images aording to the size of USPS data atered the shape desription of some etters whih ome with a very arge initia size (images of pixes). Nevertheess, N-SVM performs gobay better than OAA sine it gives ower error rates in 3 asses. Besides, it redues signifianty the error rates of asses whih are ompetey onfused by OAA (I,J, and Z). Error rate (%) Figure 3. Error rates respetive to eah ass of interest. 6. Conusions OAA N-SVM A C E G I K M O Q S U W Y Casses In this paper, a fast and robust system for handwritten aphanumeri harater reognition is proposed. Tangent simiarities based on ass speifi tangent vetors are used to dea with data variabiity whie N-SVM ombination was used to aeerate the training stage of muti-ass impementation of SVM (OAA). The N-SVM ombination takes advantage from the high separation abiity of SVM to separate the pairs of asses and the earning abiity of neura network to onstrut an automati deision about aphanumeri asses. The experimenta anaysis was onduted on a dataset obtained by ombining two benhmark datasets whih are the USPS for handwritten digits and the C-Cube for upperase etters. The resuts indiate that tangent simiarities aow a signifiant improvement in error rate when using 30 tangent vetors. Besides, it has been shown that N-SVM outperforms the OAA in both runtime and reognition auray. REFERENCES [] Kim, K. I., Jung, K., Park, S. H., and Kim, H. J., 00. Support vetor mahines for texture assifiation, IEEE Transations on Pattern Anaysis and Mahine Inteigene, Vo. 4, [] Pamondon, R., and Srihari, S. N., 000. On-ine and off-ine handwriting reognition: A omprehensive survey, IEEE Transations on Pattern Anaysis and Mahine Inteigene, Vo., [3] Burges, C. J. C., 998. A Tutoria on support vetor mahines for pattern reognition. Data Mining and Knowedge Disovery, Edited by Ussama Fayyad, Vo., -67 [4] Shökopf, B., 997. Support vetor earning, Thèse de PhD: Université de Berin, 73 pages [5] Keysers, D., Paredes, R., Ney, H., and Vida E., 00. Combination of tangent vetors and oa representations for handwritten digit reognition, Leture Notes in Computer Siene, Vo. 396, [6] Nemmour, H., and Chibani, Y., 00. Handwritten digit reognition based on a Neura-SVM ombination, to appear in Internationa journa of omputers and appiations, Vo, 3, [7] Trier, O. D., Jain, A. K., and Taxt. T., 996. Feature extration methods for harater reognition A survey, Pattern reognition, Vo. 9, [8] Nemmour, H., and Chibani, Y., 009. Handwritten aphanumeri harater reognition based on support vetor mahines and ombination of desriptors, ACM Internationa Conferene on Inteigent Computing and Information Systems ICICIS 09, 9- Marh, Cairo [9] Chen, G. Y., Bui, T. D., and Krzyzak, A., 006. Rotation invariant feature extration using ridgeet and Fourier transforms; Pattern Anaysis and Appiation Journa, Vo. 9, [0] Yang, L., Suen, C. Y., Bui, T. D., et Zhang, P., 005. Disrimination of simiar handwritten numeras based on invariant urvature features, Pattern Reognition Journa, Vo. 38, [] Broumandnia, A., Shanbehzadeh, J., and Varnoosfaderani, M. R., 008. Persian/arabi handwritten word reognition using M-band paket waveet transform, Image Vision and Computing Journa, Vo.6, [] Simard, P., Le Cun, Y., Denker, J., and Vitorri, B., 993. Effiient pattern reognition using a new transformation distane, Advanes in Neura Information Proessing Systems, Vo. 5, 50-58

6 Amerian Journa of Signa Proessing. 0; (): [3] Simard, P., Le Cun, Y., Denker, J., et Vitorri, B., 998. Transformation invariane in pattern reognition tangent distane and tangent propagation, Leture Notes on Computer Siene, Vo. 54, [4] Keysers, D., Mherey, W., and Ney, H., 004. Adaptation in statistia pattern reognition using tangent vetors, IEEE Transations on Pattern Anaysis and Mahine Inteigene, Vo. 6, [5] Vapnik, V., 995. The nature of statistia earning theory, Springer-Verag, New York [6] Hsu, C-W., and Lin, C-J., 00. A omparison of Methods for Muti-ass Support Vetor Mahines, IEEE Transations on Neura Networks, Vo. 3, [7] Joahims, T., 998. Making arge sae SVM Learning pratia, MIT Press: Advanes in Kerne Methods Support Vetor Learning, B. Sh okopf, C. J. C. Burges, and A. J. Smoa Editors, [8] D. E. Rumehart, G. E. Hinton, & R. J. Wiiams, Learning interna representations by error propagation, MIT Press (Parae Distributed Proessing: Exporations in the mirostruture of Cognition),, 986, [9] DeCoste, D., and Shökopf, B., 00. Training invariant support vetor mahines, Mahine Learning Journa, Vo. 46, 6-90 [0] Haasdonk, B., and Keysers, D., 00. Tangent distane kernes for support vetor mahines, In Pro. Of the 6 th Internationa Conferene on Pattern Reognition (ICPR), vo., pp , 00 [] Nemmour, H., and Chibani, Y., 008. Inorporating Tangent Vetors in SVM Kernes for Handwritten Digit Reognition, the th Internationa Conferene on Frontiers in Handwriting Reognition, ICFHR 08, 9- Août, Montréa, Canada

SVM: Terminology 1(6) SVM: Terminology 2(6)

SVM: Terminology 1(6) SVM: Terminology 2(6) Andrew Kusiak Inteigent Systems Laboratory 39 Seamans Center he University of Iowa Iowa City, IA 54-57 SVM he maxima margin cassifier is simiar to the perceptron: It aso assumes that the data points are

More information

Complexity of Regularization RBF Networks

Complexity of Regularization RBF Networks Complexity of Regularization RBF Networks Mark A Kon Department of Mathematis and Statistis Boston University Boston, MA 02215 mkon@buedu Leszek Plaskota Institute of Applied Mathematis University of Warsaw

More information

FOR many years the authors of this paper have worked on

FOR many years the authors of this paper have worked on The Fast Parametri Integra Equations System for Poygona D Potentia Probems Andrzej Kużeewski and Eugeniusz Zieniuk Abstrat Appiation of tehniques for modeing of boundary vaue probems impies three onfiting

More information

Genetic Algorithm Based Recurrent Fuzzy Neural Network Modeling of Chemical Processes

Genetic Algorithm Based Recurrent Fuzzy Neural Network Modeling of Chemical Processes Journa of Universa Computer Siene, vo. 3, no. 9 (27), 332-343 submitted: 2/6/6, aepted: 24//6, appeared: 28/9/7 J.UCS Geneti Agorithm Based Reurrent Fuzzy Neura Network odeing of Chemia Proesses Jii Tao,

More information

ABSOLUTELY CONTINUOUS FUNCTIONS OF TWO VARIABLES IN THE SENSE OF CARATHÉODORY

ABSOLUTELY CONTINUOUS FUNCTIONS OF TWO VARIABLES IN THE SENSE OF CARATHÉODORY Eetroni Journa of Differentia Equations, Vo. 2010(2010), No. 154, pp. 1 11. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu ftp ejde.math.txstate.edu ABSOLUTELY CONTINUOUS

More information

Sparse Structured Associative Memories as Efficient Set-Membership Data Structures

Sparse Structured Associative Memories as Efficient Set-Membership Data Structures Sparse Strutured Assoiative Memories as Effiient Set-Membership Data Strutures Vinent Gripon Eetronis Department Tééom Bretagne, Brest, Frane vinent.gripon@ens-ahan.org Vitay Skahek Institute of Computer

More information

An Algorithm for Pruning Redundant Modules in Min-Max Modular Network

An Algorithm for Pruning Redundant Modules in Min-Max Modular Network An Agorithm for Pruning Redundant Modues in Min-Max Moduar Network Hui-Cheng Lian and Bao-Liang Lu Department of Computer Science and Engineering, Shanghai Jiao Tong University 1954 Hua Shan Rd., Shanghai

More information

Research Article Some Applications of Second-Order Differential Subordination on a Class of Analytic Functions Defined by Komatu Integral Operator

Research Article Some Applications of Second-Order Differential Subordination on a Class of Analytic Functions Defined by Komatu Integral Operator ISRN Mathematia Anaysis, Artie ID 66235, 5 pages http://dx.doi.org/1.1155/214/66235 Researh Artie Some Appiations of Seond-Order Differentia Subordination on a Cass of Anayti Funtions Defined by Komatu

More information

A simple expression for radial distribution functions of pure fluids and mixtures

A simple expression for radial distribution functions of pure fluids and mixtures A simple expression for radial distribution funtions of pure fluids and mixtures Enrio Matteoli a) Istituto di Chimia Quantistia ed Energetia Moleolare, CNR, Via Risorgimento, 35, 56126 Pisa, Italy G.

More information

Bilinear Formulated Multiple Kernel Learning for Multi-class Classification Problem

Bilinear Formulated Multiple Kernel Learning for Multi-class Classification Problem Bilinear Formulated Multiple Kernel Learning for Multi-lass Classifiation Problem Takumi Kobayashi and Nobuyuki Otsu National Institute of Advaned Industrial Siene and Tehnology, -- Umezono, Tsukuba, Japan

More information

Danielle Maddix AA238 Final Project December 9, 2016

Danielle Maddix AA238 Final Project December 9, 2016 Struture and Parameter Learning in Bayesian Networks with Appliations to Prediting Breast Caner Tumor Malignany in a Lower Dimension Feature Spae Danielle Maddix AA238 Final Projet Deember 9, 2016 Abstrat

More information

SUPPLEMENTARY INFORMATION

SUPPLEMENTARY INFORMATION DOI: 10.1038/NPHOTON.013.10 Suppementary Information: Quantum teeportation using a ight emitting diode J. Nisson 1, R. M. Stevenson 1*, K. H. A. Chan 1,, J. Skiba-Szymanska 1, M. Luamarini 1, M. B. Ward

More information

Millennium Relativity Acceleration Composition. The Relativistic Relationship between Acceleration and Uniform Motion

Millennium Relativity Acceleration Composition. The Relativistic Relationship between Acceleration and Uniform Motion Millennium Relativity Aeleration Composition he Relativisti Relationship between Aeleration and niform Motion Copyright 003 Joseph A. Rybzyk Abstrat he relativisti priniples developed throughout the six

More information

Weighted K-Nearest Neighbor Revisited

Weighted K-Nearest Neighbor Revisited Weighted -Nearest Neighbor Revisited M. Biego University of Verona Verona, Italy Email: manuele.biego@univr.it M. Loog Delft University of Tehnology Delft, The Netherlands Email: m.loog@tudelft.nl Abstrat

More information

Hankel Optimal Model Order Reduction 1

Hankel Optimal Model Order Reduction 1 Massahusetts Institute of Tehnology Department of Eletrial Engineering and Computer Siene 6.245: MULTIVARIABLE CONTROL SYSTEMS by A. Megretski Hankel Optimal Model Order Redution 1 This leture overs both

More information

MINIMIZATION OF FREQUENCY-WEIGHTED l 2 -SENSITIVITY FOR MULTI-INPUT/MULTI-OUTPUT LINEAR SYSTEMS

MINIMIZATION OF FREQUENCY-WEIGHTED l 2 -SENSITIVITY FOR MULTI-INPUT/MULTI-OUTPUT LINEAR SYSTEMS Automatique et ordinateurs MINIMIZAION OF FREQUENCY-WEIGHED -SENSIIVIY FOR MULI-INPU/MULI-OUPU LINEAR SYSEMS AKAO HINAMOO, OSAMU ANAKA, AKIMISU DOI Key words: MIMO inear disrete-time systems, Frequeny-weighted

More information

SVM-based Supervised and Unsupervised Classification Schemes

SVM-based Supervised and Unsupervised Classification Schemes SVM-based Supervised and Unsupervised Cassification Schemes LUMINITA STATE University of Pitesti Facuty of Mathematics and Computer Science 1 Targu din Vae St., Pitesti 110040 ROMANIA state@cicknet.ro

More information

Uncertainty Handling CMA-ES for Reinforcement Learning

Uncertainty Handling CMA-ES for Reinforcement Learning Unertainty Handing CMA-ES for Reinforement Learning Verena Heidrih-Meisner Institut für Neuroinformatik Ruhr-Universität Bohum 44780 Bohum, Germany verena.heidrih-meisner @neuroinformatik.rub.de Christian

More information

Multilayer Kerceptron

Multilayer Kerceptron Mutiayer Kerceptron Zotán Szabó, András Lőrincz Department of Information Systems, Facuty of Informatics Eötvös Loránd University Pázmány Péter sétány 1/C H-1117, Budapest, Hungary e-mai: szzoi@csetehu,

More information

CS229 Lecture notes. Andrew Ng

CS229 Lecture notes. Andrew Ng CS229 Lecture notes Andrew Ng Part IX The EM agorithm In the previous set of notes, we taked about the EM agorithm as appied to fitting a mixture of Gaussians. In this set of notes, we give a broader view

More information

BER Performance Comparison of Overlap FDE and Sliding-window Time-domain Equalization for Single-carrier Transmission

BER Performance Comparison of Overlap FDE and Sliding-window Time-domain Equalization for Single-carrier Transmission BER Performane Comparison of Overap FDE and Siding-window ime-domain Equaization for Singe-arrier ransmission atsunori OBARA Kazuki AKEDA and Fumiyuki ADACI Dept. of Eetria and Communiation Engineering,

More information

Model-based mixture discriminant analysis an experimental study

Model-based mixture discriminant analysis an experimental study Model-based mixture disriminant analysis an experimental study Zohar Halbe and Mayer Aladjem Department of Eletrial and Computer Engineering, Ben-Gurion University of the Negev P.O.Box 653, Beer-Sheva,

More information

Nonreversibility of Multiple Unicast Networks

Nonreversibility of Multiple Unicast Networks Nonreversibility of Multiple Uniast Networks Randall Dougherty and Kenneth Zeger September 27, 2005 Abstrat We prove that for any finite direted ayli network, there exists a orresponding multiple uniast

More information

A Solution to the 4-bit Parity Problem with a Single Quaternary Neuron

A Solution to the 4-bit Parity Problem with a Single Quaternary Neuron Neura Information Processing - Letters and Reviews Vo. 5, No. 2, November 2004 LETTER A Soution to the 4-bit Parity Probem with a Singe Quaternary Neuron Tohru Nitta Nationa Institute of Advanced Industria

More information

An I-Vector Backend for Speaker Verification

An I-Vector Backend for Speaker Verification An I-Vetor Bakend for Speaker Verifiation Patrik Kenny, 1 Themos Stafylakis, 1 Jahangir Alam, 1 and Marel Kokmann 2 1 CRIM, Canada, {patrik.kenny, themos.stafylakis, jahangir.alam}@rim.a 2 VoieTrust, Canada,

More information

Strong Tracking Unscented Kalman Filtering Algorithm Based-on Satellite Attitude Determination System

Strong Tracking Unscented Kalman Filtering Algorithm Based-on Satellite Attitude Determination System Internationa Journa of Future Generation Communiation and Networing Vo.7, No.3 (, pp.55-66 http://dx.doi.org/.57/ijfgn..7.3. Strong raing Unsented Kaman Fitering Agorithm Based-on Sateite Attitude Determination

More information

How the backpropagation algorithm works Srikumar Ramalingam School of Computing University of Utah

How the backpropagation algorithm works Srikumar Ramalingam School of Computing University of Utah How the backpropagation agorithm works Srikumar Ramaingam Schoo of Computing University of Utah Reference Most of the sides are taken from the second chapter of the onine book by Michae Nieson: neuranetworksanddeepearning.com

More information

Batch Process Monitoring Using Two-Dimensional Hidden Semi-Markov Models

Batch Process Monitoring Using Two-Dimensional Hidden Semi-Markov Models Bath Proess Monitoring Using wo-imensiona Hidden Semi-Marov Modes J. Chen, Member, IAENG and Y.-C. Jiang Abstrat In this paper, a nove monitoring method for the repetitive bath operation with two-dimensiona

More information

A Unified View on Multi-class Support Vector Classification Supplement

A Unified View on Multi-class Support Vector Classification Supplement Journal of Mahine Learning Researh??) Submitted 7/15; Published?/?? A Unified View on Multi-lass Support Vetor Classifiation Supplement Ürün Doğan Mirosoft Researh Tobias Glasmahers Institut für Neuroinformatik

More information

Advanced Computational Fluid Dynamics AA215A Lecture 4

Advanced Computational Fluid Dynamics AA215A Lecture 4 Advaned Computational Fluid Dynamis AA5A Leture 4 Antony Jameson Winter Quarter,, Stanford, CA Abstrat Leture 4 overs analysis of the equations of gas dynamis Contents Analysis of the equations of gas

More information

Assessing the Performance of a BCI: A Task-Oriented Approach

Assessing the Performance of a BCI: A Task-Oriented Approach Assessing the Performane of a BCI: A Task-Oriented Approah B. Dal Seno, L. Mainardi 2, M. Matteui Department of Eletronis and Information, IIT-Unit, Politenio di Milano, Italy 2 Department of Bioengineering,

More information

Sensitivity Analysis in Markov Networks

Sensitivity Analysis in Markov Networks Sensitivity Analysis in Markov Networks Hei Chan and Adnan Darwihe Computer Siene Department University of California, Los Angeles Los Angeles, CA 90095 {hei,darwihe}@s.ula.edu Abstrat This paper explores

More information

An Integrated Architecture of Adaptive Neural Network Control for Dynamic Systems

An Integrated Architecture of Adaptive Neural Network Control for Dynamic Systems An Integrated Arhiteture of Adaptive Neural Network Control for Dynami Systems Robert L. Tokar 2 Brian D.MVey2 'Center for Nonlinear Studies, 2Applied Theoretial Physis Division Los Alamos National Laboratory,

More information

A Spatiotemporal Approach to Passive Sound Source Localization

A Spatiotemporal Approach to Passive Sound Source Localization A Spatiotemporal Approah Passive Sound Soure Loalization Pasi Pertilä, Mikko Parviainen, Teemu Korhonen and Ari Visa Institute of Signal Proessing Tampere University of Tehnology, P.O.Box 553, FIN-330,

More information

Error Bounds for Context Reduction and Feature Omission

Error Bounds for Context Reduction and Feature Omission Error Bounds for Context Redution and Feature Omission Eugen Bek, Ralf Shlüter, Hermann Ney,2 Human Language Tehnology and Pattern Reognition, Computer Siene Department RWTH Aahen University, Ahornstr.

More information

Statistical Learning Theory: A Primer

Statistical Learning Theory: A Primer Internationa Journa of Computer Vision 38(), 9 3, 2000 c 2000 uwer Academic Pubishers. Manufactured in The Netherands. Statistica Learning Theory: A Primer THEODOROS EVGENIOU, MASSIMILIANO PONTIL AND TOMASO

More information

MUSIC GENRE CLASSIFICATION USING LOCALITY PRESERVING NON-NEGATIVE TENSOR FACTORIZATION AND SPARSE REPRESENTATIONS

MUSIC GENRE CLASSIFICATION USING LOCALITY PRESERVING NON-NEGATIVE TENSOR FACTORIZATION AND SPARSE REPRESENTATIONS 10th International Soiety for Musi Information Retrieval Conferene (ISMIR 2009) MUSIC GENRE CLASSIFICATION USING LOCALITY PRESERVING NON-NEGATIVE TENSOR FACTORIZATION AND SPARSE REPRESENTATIONS Yannis

More information

Constructing Gyro-free Inertial Measurement Unit from Dual Accelerometers for Gesture Detection

Constructing Gyro-free Inertial Measurement Unit from Dual Accelerometers for Gesture Detection Sensors & ransduers Vo. 7 Issue 5 May 04 pp. 34-40 Sensors & ransduers 04 by IFSA Pubishing S. L. http://www.sensorsporta.om Construting Gyro-free Inertia Measurement Unit from Dua Aeerometers for Gesture

More information

New Algorithms for Nonlinear Generalized Disjunctive Programming

New Algorithms for Nonlinear Generalized Disjunctive Programming ew Agorithms for oninear Generaized Disuntive Programming Sangbum Lee and Ignaio E. Grossmann * Department of Chemia Engineering Carnegie Meon University Pittsburgh PA U.S.A. Otober 999 / Marh * To whom

More information

Optimization of Statistical Decisions for Age Replacement Problems via a New Pivotal Quantity Averaging Approach

Optimization of Statistical Decisions for Age Replacement Problems via a New Pivotal Quantity Averaging Approach Amerian Journal of heoretial and Applied tatistis 6; 5(-): -8 Published online January 7, 6 (http://www.sienepublishinggroup.om/j/ajtas) doi:.648/j.ajtas.s.65.4 IN: 36-8999 (Print); IN: 36-96 (Online)

More information

Variation Based Online Travel Time Prediction Using Clustered Neural Networks

Variation Based Online Travel Time Prediction Using Clustered Neural Networks Variation Based Online Travel Time Predition Using lustered Neural Networks Jie Yu, Gang-Len hang, H.W. Ho and Yue Liu Abstrat-This paper proposes a variation-based online travel time predition approah

More information

Feature Selection by Independent Component Analysis and Mutual Information Maximization in EEG Signal Classification

Feature Selection by Independent Component Analysis and Mutual Information Maximization in EEG Signal Classification Feature Seletion by Independent Component Analysis and Mutual Information Maximization in EEG Signal Classifiation Tian Lan, Deniz Erdogmus, Andre Adami, Mihael Pavel BME Department, Oregon Health & Siene

More information

Using the Green s Function to find the Solution to the Wave. Equation:

Using the Green s Function to find the Solution to the Wave. Equation: Using the Green s Funtion to find the Soution to the Wave Exampe 1: 2 1 2 2 t 2 Equation: r,t q 0 e it r aẑ r aẑ r,t r 1 r ; r r,t r 1 r 2 The Green s funtion soution is given by r,t G R r r,t t Fr,t d

More information

Stochastic Variational Inference with Gradient Linearization

Stochastic Variational Inference with Gradient Linearization Stochastic Variationa Inference with Gradient Linearization Suppementa Materia Tobias Pötz * Anne S Wannenwetsch Stefan Roth Department of Computer Science, TU Darmstadt Preface In this suppementa materia,

More information

Time-varying Stiffness Characteristics of Shaft with Slant Crack

Time-varying Stiffness Characteristics of Shaft with Slant Crack Internationa Conferene on Modeing, Simuation and Appied Mathematis (MSAM 05) Time-varying Stiffness Charateristis of Shaft with Sant Cra Hengheng Xia Shoo of Aeronautia Manufaturing Engineering Nanhang

More information

How the backpropagation algorithm works Srikumar Ramalingam School of Computing University of Utah

How the backpropagation algorithm works Srikumar Ramalingam School of Computing University of Utah How the backpropagation agorithm works Srikumar Ramaingam Schoo of Computing University of Utah Reference Most of the sides are taken from the second chapter of the onine book by Michae Nieson: neuranetworksanddeepearning.com

More information

Naïve Bayes for Text Classification

Naïve Bayes for Text Classification Naïve Bayes for Text Cassifiation adapted by Lye Ungar from sides by Mith Marus, whih were adapted from sides by Massimo Poesio, whih were adapted from sides by Chris Manning : Exampe: Is this spam? From:

More information

7 Max-Flow Problems. Business Computing and Operations Research 608

7 Max-Flow Problems. Business Computing and Operations Research 608 7 Max-Flow Problems Business Computing and Operations Researh 68 7. Max-Flow Problems In what follows, we onsider a somewhat modified problem onstellation Instead of osts of transmission, vetor now indiates

More information

10.5 Unsupervised Bayesian Learning

10.5 Unsupervised Bayesian Learning The Bayes Classifier Maximum-likelihood methods: Li Yu Hongda Mao Joan Wang parameter vetor is a fixed but unknown value Bayes methods: parameter vetor is a random variable with known prior distribution

More information

Wavetech, LLC. Ultrafast Pulses and GVD. John O Hara Created: Dec. 6, 2013

Wavetech, LLC. Ultrafast Pulses and GVD. John O Hara Created: Dec. 6, 2013 Ultrafast Pulses and GVD John O Hara Created: De. 6, 3 Introdution This doument overs the basi onepts of group veloity dispersion (GVD) and ultrafast pulse propagation in an optial fiber. Neessarily, it

More information

Singular Event Detection

Singular Event Detection Singular Event Detetion Rafael S. Garía Eletrial Engineering University of Puerto Rio at Mayagüez Rafael.Garia@ee.uprm.edu Faulty Mentor: S. Shankar Sastry Researh Supervisor: Jonathan Sprinkle Graduate

More information

Intuitionistic Fuzzy Set and Its Application in Selecting Specialization: A Case Study for Engineering Students

Intuitionistic Fuzzy Set and Its Application in Selecting Specialization: A Case Study for Engineering Students International Journal of Mathematial nalysis and ppliations 2015; 2(6): 74-78 Published online Deember 17, 2015 (http://www.aasit.org/journal/ijmaa) ISSN: 2375-3927 Intuitionisti Fuzzy Set and Its ppliation

More information

Lecture 3 - Lorentz Transformations

Lecture 3 - Lorentz Transformations Leture - Lorentz Transformations A Puzzle... Example A ruler is positioned perpendiular to a wall. A stik of length L flies by at speed v. It travels in front of the ruler, so that it obsures part of the

More information

Annealing Chaotic Pattern Search Learning Method for Multilayer Neural Networks

Annealing Chaotic Pattern Search Learning Method for Multilayer Neural Networks IJCSNS Internationa Journa of Computer Siene and Networ Seurity, VOL.7 No.6, June 2007 137 Anneaing Chaoti Pattern Searh Learning Method for Mutiayer Neura Networs Shange Gao, Hongwei Dai, Yunyi Zhu and

More information

The research of Q-ary LDPC codes in the navigation system

The research of Q-ary LDPC codes in the navigation system 3rd Internationa Conferene on Mehatronis and Information Tehnoogy (ICMIT 26) The researh of Q-ary LDPC odes in the naigation system Xu Yi, Daoxing Guo2 and Wei You3 Coege of Communiations Engineering,

More information

COMBINED PROBE FOR MACH NUMBER, TEMPERATURE AND INCIDENCE INDICATION

COMBINED PROBE FOR MACH NUMBER, TEMPERATURE AND INCIDENCE INDICATION 4 TH INTERNATIONAL CONGRESS OF THE AERONAUTICAL SCIENCES COMBINED PROBE FOR MACH NUMBER, TEMPERATURE AND INCIDENCE INDICATION Jiri Nozika*, Josef Adame*, Daniel Hanus** *Department of Fluid Dynamis and

More information

Developing Excel Macros for Solving Heat Diffusion Problems

Developing Excel Macros for Solving Heat Diffusion Problems Session 50 Developing Exel Maros for Solving Heat Diffusion Problems N. N. Sarker and M. A. Ketkar Department of Engineering Tehnology Prairie View A&M University Prairie View, TX 77446 Abstrat This paper

More information

A NETWORK SIMPLEX ALGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM

A NETWORK SIMPLEX ALGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM NETWORK SIMPLEX LGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM Cen Çalışan, Utah Valley University, 800 W. University Parway, Orem, UT 84058, 801-863-6487, en.alisan@uvu.edu BSTRCT The minimum

More information

Einstein s Three Mistakes in Special Relativity Revealed. Copyright Joseph A. Rybczyk

Einstein s Three Mistakes in Special Relativity Revealed. Copyright Joseph A. Rybczyk Einstein s Three Mistakes in Speial Relativity Revealed Copyright Joseph A. Rybzyk Abstrat When the evidene supported priniples of eletromagneti propagation are properly applied, the derived theory is

More information

A Small Footprint i-vector Extractor

A Small Footprint i-vector Extractor A Small Footprint i-vetor Extrator Patrik Kenny Centre de reherhe informatique de Montréal (CRIM) Patrik.Kenny@rim.a Abstrat Both the memory and omputational requirements of algorithms traditionally used

More information

Time Domain Method of Moments

Time Domain Method of Moments Time Domain Method of Moments Massahusetts Institute of Tehnology 6.635 leture notes 1 Introdution The Method of Moments (MoM) introdued in the previous leture is widely used for solving integral equations

More information

A new method of measuring similarity between two neutrosophic soft sets and its application in pattern recognition problems

A new method of measuring similarity between two neutrosophic soft sets and its application in pattern recognition problems Neutrosophi Sets and Systems, Vol. 8, 05 63 A new method of measuring similarity between two neutrosophi soft sets and its appliation in pattern reognition problems Anjan Mukherjee, Sadhan Sarkar, Department

More information

EXACT TRAVELLING WAVE SOLUTIONS FOR THE GENERALIZED KURAMOTO-SIVASHINSKY EQUATION

EXACT TRAVELLING WAVE SOLUTIONS FOR THE GENERALIZED KURAMOTO-SIVASHINSKY EQUATION Journal of Mathematial Sienes: Advanes and Appliations Volume 3, 05, Pages -3 EXACT TRAVELLING WAVE SOLUTIONS FOR THE GENERALIZED KURAMOTO-SIVASHINSKY EQUATION JIAN YANG, XIAOJUAN LU and SHENGQIANG TANG

More information

Counting Idempotent Relations

Counting Idempotent Relations Counting Idempotent Relations Beriht-Nr. 2008-15 Florian Kammüller ISSN 1436-9915 2 Abstrat This artile introdues and motivates idempotent relations. It summarizes haraterizations of idempotents and their

More information

A NONLILEAR CONTROLLER FOR SHIP AUTOPILOTS

A NONLILEAR CONTROLLER FOR SHIP AUTOPILOTS Vietnam Journal of Mehanis, VAST, Vol. 4, No. (), pp. A NONLILEAR CONTROLLER FOR SHIP AUTOPILOTS Le Thanh Tung Hanoi University of Siene and Tehnology, Vietnam Abstrat. Conventional ship autopilots are

More information

Available online at ScienceDirect. Energy Procedia 76 (2015 ) European Geosciences Union General Assembly 2015, EGU

Available online at  ScienceDirect. Energy Procedia 76 (2015 ) European Geosciences Union General Assembly 2015, EGU Avaiabe onine at www.sienediret.om SieneDiret Energy Proedia 76 (05 ) 406 4 European Geosienes Union Genera Assemby 05, EGU Division Energy, Resoures & Environment, ERE Appiation of stoasti metods to doube

More information

A Queueing Model for Call Blending in Call Centers

A Queueing Model for Call Blending in Call Centers A Queueing Model for Call Blending in Call Centers Sandjai Bhulai and Ger Koole Vrije Universiteit Amsterdam Faulty of Sienes De Boelelaan 1081a 1081 HV Amsterdam The Netherlands E-mail: {sbhulai, koole}@s.vu.nl

More information

Evaluation of effect of blade internal modes on sensitivity of Advanced LIGO

Evaluation of effect of blade internal modes on sensitivity of Advanced LIGO Evaluation of effet of blade internal modes on sensitivity of Advaned LIGO T0074-00-R Norna A Robertson 5 th Otober 00. Introdution The urrent model used to estimate the isolation ahieved by the quadruple

More information

Aircraft CAS Design with Input Saturation Using Dynamic Model Inversion

Aircraft CAS Design with Input Saturation Using Dynamic Model Inversion International Journal of Control, Automation, and Systems Vol., No. 3, September 003 35 Airraft CAS Design with Input Saturation sing Dynami Model Inversion Sangsoo Lim and Byoung Soo Kim Abstrat: This

More information

Supplementary Materials

Supplementary Materials Supplementary Materials Neural population partitioning and a onurrent brain-mahine interfae for sequential motor funtion Maryam M. Shanehi, Rollin C. Hu, Marissa Powers, Gregory W. Wornell, Emery N. Brown

More information

The Effectiveness of the Linear Hull Effect

The Effectiveness of the Linear Hull Effect The Effetiveness of the Linear Hull Effet S. Murphy Tehnial Report RHUL MA 009 9 6 Otober 009 Department of Mathematis Royal Holloway, University of London Egham, Surrey TW0 0EX, England http://www.rhul.a.uk/mathematis/tehreports

More information

arxiv: v2 [cond-mat.dis-nn] 29 Mar 2017

arxiv: v2 [cond-mat.dis-nn] 29 Mar 2017 arxiv:69.8375v [ond-mat.dis-nn] 9 Mar 7 The distribution of first hitting times of non-baktraking random waks on Erdős-Rényi networks Ido Tishby, Ofer Biham and Eytan Katzav Raah Institute of Physis, The

More information

Measuring & Inducing Neural Activity Using Extracellular Fields I: Inverse systems approach

Measuring & Inducing Neural Activity Using Extracellular Fields I: Inverse systems approach Measuring & Induing Neural Ativity Using Extraellular Fields I: Inverse systems approah Keith Dillon Department of Eletrial and Computer Engineering University of California San Diego 9500 Gilman Dr. La

More information

Parametric and sensitivity analysis of a vibratory automobile model

Parametric and sensitivity analysis of a vibratory automobile model Louisiana State University LSU Digita Commons LSU Master's Theses Graduate Shoo Parametri and sensitivity anaysis of a vibratory automobie mode Kania Nioe Vesse Louisiana State University and Agriutura

More information

Determining The Degree of Generalization Using An Incremental Learning Algorithm

Determining The Degree of Generalization Using An Incremental Learning Algorithm Determining The Degree of Generaization Using An Incrementa Learning Agorithm Pabo Zegers Facutad de Ingeniería, Universidad de os Andes San Caros de Apoquindo 22, Las Condes, Santiago, Chie pzegers@uandes.c

More information

Advances in Radio Science

Advances in Radio Science Advanes in adio Siene 2003) 1: 99 104 Copernius GmbH 2003 Advanes in adio Siene A hybrid method ombining the FDTD and a time domain boundary-integral equation marhing-on-in-time algorithm A Beker and V

More information

Relative Maxima and Minima sections 4.3

Relative Maxima and Minima sections 4.3 Relative Maxima and Minima setions 4.3 Definition. By a ritial point of a funtion f we mean a point x 0 in the domain at whih either the derivative is zero or it does not exists. So, geometrially, one

More information

QCLAS Sensor for Purity Monitoring in Medical Gas Supply Lines

QCLAS Sensor for Purity Monitoring in Medical Gas Supply Lines DOI.56/sensoren6/P3. QLAS Sensor for Purity Monitoring in Medial Gas Supply Lines Henrik Zimmermann, Mathias Wiese, Alessandro Ragnoni neoplas ontrol GmbH, Walther-Rathenau-Str. 49a, 7489 Greifswald, Germany

More information

Computer Science 786S - Statistical Methods in Natural Language Processing and Data Analysis Page 1

Computer Science 786S - Statistical Methods in Natural Language Processing and Data Analysis Page 1 Computer Siene 786S - Statistial Methods in Natural Language Proessing and Data Analysis Page 1 Hypothesis Testing A statistial hypothesis is a statement about the nature of the distribution of a random

More information

A Differential Equation for Specific Catchment Area

A Differential Equation for Specific Catchment Area Proeedings of Geomorphometry 2009. Zurih, Sitzerland, 3 ugust - 2 September, 2009 Differential Equation for Speifi Cathment rea J. C. Gallant, M. F. Huthinson 2 CSIRO Land and Water, GPO Box 666, Canberra

More information

Keywords: Controllers with PI and PID dynamics, Takagi-Sugeno fuzzy models, Takagi- Sugeno fuzzy controllers, stability analysis, winding system.

Keywords: Controllers with PI and PID dynamics, Takagi-Sugeno fuzzy models, Takagi- Sugeno fuzzy controllers, stability analysis, winding system. Ata Poytehnia Hungaria Vo. 2, No., 2005 Deveopment of Conventiona and Fuzzy Controers and Takagi-Sugeno Fuzzy Modes Dediated for Contro of Low Order Benhmarks with Time Variabe Parameters Stefan Preit

More information

Lightpath routing for maximum reliability in optical mesh networks

Lightpath routing for maximum reliability in optical mesh networks Vol. 7, No. 5 / May 2008 / JOURNAL OF OPTICAL NETWORKING 449 Lightpath routing for maximum reliability in optial mesh networks Shengli Yuan, 1, * Saket Varma, 2 and Jason P. Jue 2 1 Department of Computer

More information

ONLINE APPENDICES for Cost-Effective Quality Assurance in Crowd Labeling

ONLINE APPENDICES for Cost-Effective Quality Assurance in Crowd Labeling ONLINE APPENDICES for Cost-Effetive Quality Assurane in Crowd Labeling Jing Wang Shool of Business and Management Hong Kong University of Siene and Tehnology Clear Water Bay Kowloon Hong Kong jwang@usthk

More information

Expectation-Maximization for Estimating Parameters for a Mixture of Poissons

Expectation-Maximization for Estimating Parameters for a Mixture of Poissons Expectation-Maximization for Estimating Parameters for a Mixture of Poissons Brandon Maone Department of Computer Science University of Hesini February 18, 2014 Abstract This document derives, in excrutiating

More information

Robust Flight Control Design for a Turn Coordination System with Parameter Uncertainties

Robust Flight Control Design for a Turn Coordination System with Parameter Uncertainties Amerian Journal of Applied Sienes 4 (7): 496-501, 007 ISSN 1546-939 007 Siene Publiations Robust Flight ontrol Design for a urn oordination System with Parameter Unertainties 1 Ari Legowo and Hiroshi Okubo

More information

BP neural network-based sports performance prediction model applied research

BP neural network-based sports performance prediction model applied research Avaiabe onine www.jocpr.com Journa of Chemica and Pharmaceutica Research, 204, 6(7:93-936 Research Artice ISSN : 0975-7384 CODEN(USA : JCPRC5 BP neura networ-based sports performance prediction mode appied

More information

Simplified Buckling Analysis of Skeletal Structures

Simplified Buckling Analysis of Skeletal Structures Simplified Bukling Analysis of Skeletal Strutures B.A. Izzuddin 1 ABSRAC A simplified approah is proposed for bukling analysis of skeletal strutures, whih employs a rotational spring analogy for the formulation

More information

Word of Mass: The Relationship between Mass Media and Word-of-Mouth

Word of Mass: The Relationship between Mass Media and Word-of-Mouth Word of Mass: The Relationship between Mass Media and Word-of-Mouth Roman Chuhay Preliminary version Marh 6, 015 Abstrat This paper studies the optimal priing and advertising strategies of a firm in the

More information

Coding for Random Projections and Approximate Near Neighbor Search

Coding for Random Projections and Approximate Near Neighbor Search Coding for Random Projetions and Approximate Near Neighbor Searh Ping Li Department of Statistis & Biostatistis Department of Computer Siene Rutgers University Pisataay, NJ 8854, USA pingli@stat.rutgers.edu

More information

Chapter 13, Chemical Equilibrium

Chapter 13, Chemical Equilibrium Chapter 13, Chemial Equilibrium You may have gotten the impression that when 2 reatants mix, the ensuing rxn goes to ompletion. In other words, reatants are onverted ompletely to produts. We will now learn

More information

A Combination of BIM and BEM for efficiently analyzing optical. elements

A Combination of BIM and BEM for efficiently analyzing optical. elements A Combination of BIM and BEM for effiiently analyzing optial elements SUN Fang, LIU Juan, ZHANG Guo-ting, HU Chuan-fei, SU Xiaoxing Institute of Lightwave Tehnology, hool of eletronis and information engineering,

More information

LOGISTIC REGRESSION IN DEPRESSION CLASSIFICATION

LOGISTIC REGRESSION IN DEPRESSION CLASSIFICATION LOGISIC REGRESSIO I DEPRESSIO CLASSIFICAIO J. Kual,. V. ran, M. Bareš KSE, FJFI, CVU v Praze PCP, CS, 3LF UK v Praze Abstrat Well nown logisti regression and the other binary response models an be used

More information

Optimality of Inference in Hierarchical Coding for Distributed Object-Based Representations

Optimality of Inference in Hierarchical Coding for Distributed Object-Based Representations Optimaity of Inference in Hierarchica Coding for Distributed Object-Based Representations Simon Brodeur, Jean Rouat NECOTIS, Département génie éectrique et génie informatique, Université de Sherbrooke,

More information

Applying CIECAM02 for Mobile Display Viewing Conditions

Applying CIECAM02 for Mobile Display Viewing Conditions Applying CIECAM2 for Mobile Display Viewing Conditions YungKyung Park*, ChangJun Li*, M.. Luo*, Youngshin Kwak**, Du-Sik Park **, and Changyeong Kim**; * University of Leeds, Colour Imaging Lab, UK*, **

More information

On the Licensing of Innovations under Strategic Delegation

On the Licensing of Innovations under Strategic Delegation On the Liensing of Innovations under Strategi Delegation Judy Hsu Institute of Finanial Management Nanhua University Taiwan and X. Henry Wang Department of Eonomis University of Missouri USA Abstrat This

More information

The Square of the Dirichlet-to-Neumann map equals minus Laplacian

The Square of the Dirichlet-to-Neumann map equals minus Laplacian The Square of the Dirihet-to-Neumann map equas minus Lapaian D V Ingerman Abstrat. The Dirihet-to-Neumann maps onnet boundary vaues of harmoni funtions. It is an amazing fat that the square of the non-oa

More information

Control Theory association of mathematics and engineering

Control Theory association of mathematics and engineering Control Theory assoiation of mathematis and engineering Wojieh Mitkowski Krzysztof Oprzedkiewiz Department of Automatis AGH Univ. of Siene & Tehnology, Craow, Poland, Abstrat In this paper a methodology

More information

11.1 Polynomial Least-Squares Curve Fit

11.1 Polynomial Least-Squares Curve Fit 11.1 Polynomial Least-Squares Curve Fit A. Purpose This subroutine determines a univariate polynomial that fits a given disrete set of data in the sense of minimizing the weighted sum of squares of residuals.

More information

FRST Multivariate Statistics. Multivariate Discriminant Analysis (MDA)

FRST Multivariate Statistics. Multivariate Discriminant Analysis (MDA) 1 FRST 531 -- Mutivariate Statistics Mutivariate Discriminant Anaysis (MDA) Purpose: 1. To predict which group (Y) an observation beongs to based on the characteristics of p predictor (X) variabes, using

More information

From Margins to Probabilities in Multiclass Learning Problems

From Margins to Probabilities in Multiclass Learning Problems From Margins to Probabiities in Muticass Learning Probems Andrea Passerini and Massimiiano Ponti 2 and Paoo Frasconi 3 Abstract. We study the probem of muticass cassification within the framework of error

More information