Wavelet Domain Approximate Entropy-Based Epileptic Seizure Detection
|
|
- Edith Jennings
- 5 years ago
- Views:
Transcription
1 Wavelet Doman Approxmate Entropy-Based Epleptc Sezure Detecton A.S. Muthanantha Murugavel #1, S. Ramakrshnan *2 # Department of Informaton Technology Dr.Mahalngam College of Engneerng and Technology Pollach , Tamlnadu, Inda 1 as_ananth2k1@yahoo.com 2 ram_f77@yahoo.com Abstract The electroencephalogram (EEG) sgnal plays an mportant role n the detecton of eplepsy. The EEG recordngs of the ambulatory recordng systems generate very lengthy data and the detecton of the epleptc actvty requres a tmeconsumng analyss of the entre length of the EEG data by an expert. The am of ths work s to develop a new method for automatc detecton of EEG patterns usng wavelet based approxmate entropy (ApEn) and probablstc neural network (PNN). Our method conssts of EEG data collecton, feature extracton and classfcaton stages. ApEn s a statstcal parameter that measures the predctablty of the current ampltude values of a physologcal sgnal based on ts prevous ampltude values. In feature extracton stage we use best bass mother wavelet functons and wavelet thresholdng technque. For the feature selecton we have used a new methodology, that s mnmal varance wthn class and maxmal absolute dfference between classes are used for feature selecton. In classfcaton stage we mplement PNN to detect epleptc sezure detecton. It s known that the value of the ApEn drops sharply durng an epleptc sezure and ths fact s used n the proposed system and overall accuraces as hgh as 100% can be acheved by usng the proposed system.. Keywords approxmate entropy (ApEn), wavelet transform, artfcal neural network (ANN), electroencephalogram (EEG), EEG classfcaton, eplepsy, sezure detecton, probablstc neural network (PNN). I. INTRODUCTION Eplepsy s a chronc dsorder characterzed by recurrent sezures whch may vary from muscle jerks to several convolutons. Estmated 1% of world populaton suffers from eplepsy [1], whle 85% of them lve n the developng countres. Epleptc detecton s done from electroencephalogram (EEG) sgnal as eplepsy s a condton related to the bran s electrcal actvty. Electroencephalogram (EEG) s routnely used clncally to dagnose, montor and localze epleptogenc zone. Occurrence of recurrent sezures n the EEG sgnal s characterstcs of eplepsy. In majorty of the cases, the onset of the sezures cannot be predcted n a short perod, a contnuous recordng of the EEG s requred to detect eplepsy. The entre length of the EEG recordngs s analyzed by expert to detect the traces of eplepsy. The tradtonal methods of analyss are tedous and tmeconsumng and so many automated epleptc EEG detecton systems have been developed [2]. Ths paper dscusses an automated epleptc EEG detecton system usng probablstc neural network (PNN) usng a tme-frequency doman feature of the EEG sgnal called approxmate entropy (ApEn). EEG data s frst dgtzed. The dgtal EEG data s fed as an nput to an automated sezure detecton system n order to detect the sezures present n the EEG data. Approxmate Entropy drops abruptly due to the synchronous dscharge of large groups of neurons durng an epleptc actvty. Hence, t s a good feature to make use of n the automated detecton of eplepsy. Entropy s a thermodynamc quantty descrbng the amount of dsorder n the system. From an nformaton theory perspectve, the above concept of entropy s generalzed as the amount of nformaton stored n a more general probablty dstrbuton. Frst Shannon appled the concept of nformaton or logcal entropy to the scence of nformaton theory and data communcatons. Recently, a number of dfferent entropy estmators [2] have been appled to quantfy the complexty of the sgnal. Entropy estmators are broadly classfed nto two categores spectral entropes and embeddng entropes. The spectral entropes use the ampltude components of the power spectrum of the sgnal as the probabltes n entropy calculatons. It quantfes the spectral complexty of the tme seres. The embeddng entropes use the tme seres drectly to estmate the entropy. Kolmogorov Sna entropy and the approxmate entropy are the embeddng entropes dscussed here [3]. The dscrete wavelet transform s a versatle sgnal processng tool that has many engneerng and scentfc applcatons. DWT employs two sets of functons called scalng functons and wavelet functons, whch are assocated wth low-pass and hgh-pass flters, respectvely. The decomposton of the sgnal nto the dfferent frequency bands s smply obtaned by successve hgh-pass and low pass flterng of the tme doman sgnal. Subas [4] deals wth a novel method of analyss of EEG sgnals usng dscrete wavelet transform, and classfcaton usng ANN. The pseudo Wgner-Vlle and the smoothed pseudo Wgner-Vlle
2 dstrbuton [5] was used for extractng features from the tmefrequency plane. PNN s predomnantly a classfer snce t can map any nput pattern to a number of classfcatons. Among the man advantages that dscrmnate PNN s: Fast tranng process, an nherently parallel structure, guaranteed to converge to an optmal classfer as the sze of the epresentatve tranng set ncreases and tranng samples can be added or removed wthout extensve retranng. Accordngly, a PNN learns more quckly than many neural networks model, whch led to ts success on varety of applcatons. Based on these facts and advantages, PNN can be vewed as a supervsed neural network that s capable of usng t n system classfcaton and pattern recognton [6]. The detecton of eplepsy, whch ncludes vsual scannng of EEG recordngs for the spkes and sezures, s very tme consumng, especally n the case of long recordngs. In addton, bo-sgnals are hghly subjectve so dsagreement on the same record s possble, so the EEG sgnal parameters extracted and analyzed usng computers, are hghly useful n dagnostcs. Automatc analyss of EEG recordngs n the dagnoss of eplepsy started n the early 1970s and lot of sezure detecton algorthms have been developed. In ths paper, approxmate entropy-based epleptc EEG detecton proposed by the author Varavan Srnvasan et al [11], s used wth some modfed approach such as wavelet doman. Snce wavelet has several advantages, t s both tme and frequency based and t can smultaneously possess compact support, orthogonalty, symmetry, and short support, and hgh order approxmaton. Therefore, man objectve of ths paper s to propose a novel feature extracton technque for the detecton of eplepsy. The wavelet transformaton s used for extractng Approxmate Entropy and a new methodology s presented for feature selecton. The methodology s appled to two dfferent groups of EEG sgnals for analyss of EEGs and EEG sub bands for detecton of epleptc sezure: 1) healthy subjects; 2) epleptc subjects durng a sezure (ctal EEG). Each EEG s decomposed nto two consttuent EEG sub bands: delta, theta, alpha, beta, and gamma usng wavelet-based flters. The features such as Approxmate Entropy of the wavelet coeffcents are used to represent the tme frequency dstrbuton of the EEG sgnals n each sub-band of the wavelet transformaton and the probablstc neural network s used to detect epleptc EEG sgnals. II. PROPOSED METHODOLOGY As n tradtonal pattern recognton systems, the epleptc sezure detecton conssts of man modules such as a feature extractor that generates a wavelet based feature from the EEG sgnals, feature selecton that composes composte features, and a feature classfer (PNN) that outputs the class based on the composte features. The data flow of the proposed approach s llustrated n Fg. 1. Test EEG Sgnal Fg. 1 Data flow dagram of the proposed system A. Dataset Descrpton Input Sgnals (EEG Sgnals) Wavelet Transformaton Feature Extracton (ApEn) Feature Selecton Traned Probablstc Neural Network Epleptc Detecton The data used n ths research are a subset of the EEG data for both healthy and epleptc subjects made avalable onlne by Dr. Ralph Andrzejak of the Eplepsy Centre at the Unversty of Bonn, Germany ( data.html) [1]. EEGs from two dfferent groups: group H (healthy subjects) and group S (epleptc subjects durng sezure) are analyzed. The type of eplepsy was dagnosed as temporal lobe eplepsy wth the epleptogenc focus beng the hppocampal formaton. Each group contans 100 sngle channel EEG segments of 23.6 sec duraton each sampled at Hz. As such, each data segment contans N=4097 data ponts collected at ntervals of 1/173.61th of 1s. Each EEG segment s consdered as a separate EEG sgnal resultng n a total of 200 EEG sgnals or EEGs. As an example, the frst 6s of two EEGs (sgnal numbers n parentheses) for groups H (H029) and S (S001) are magnfed and dsplayed n Fg. 2. Fg. 2 Sample unfltered EEGs (0 6 s) for (from top to bottom) Group H (H029) and Group S (S001)
3 B. Wavelet Transformaton Wavelet transform s a spectral estmaton technque n whch any general functon can be expressed as an nfnte seres of wavelets. The basc dea underlyng wavelet analyss conssts of expressng a sgnal as a lnear combnaton of a partcular set of functons (wavelet transform, WT), obtaned by shftng and dlatng one sngle functon called a mother wavelet. The decomposton of the sgnal leads to a set of coeffcents called wavelet coeffcents. Therefore the sgnal can be reconstructed as a lnear combnaton of the wavelet functons weghted by the wavelet coeffcents. The key feature of wavelets s the tme-frequency localzaton. It means that most of the energy of the wavelet s restrcted to a fnte tme nterval. The wavelet technque appled to the EEG sgnal wll reveal features related to the transent nature of the sgnal, whch s not made obvous by the Fourer transform. Adel et al. [7] gave an overvew of the dscrete wavelet transform (DWT) developed for recognzng and quantfyng spkes, sharp waves and spke-waves. In general, t must be sad that no tme-frequency regons but rather tme-scale regons are defned. All wavelet transforms can be specfed n terms of a low-pass flter, whch satsfes the standard quadrature mrror flter condton. One area n whch the wavelet transformaton has been partcularly successful s the epleptc sezure detecton because t captures transent features and localzes them n both tme and frequency content accurately. The wavelet transformaton analyses the sgnal at dfferent frequency bands, wth dfferent resolutons by decomposng the sgnal nto a coarse approxmaton and detal nformaton [8]. The decomposton of the sgnal nto the dfferent frequency bands s merely obtaned by consecutve hgh-pass and low-pass flterng of the tme doman sgnal. The procedure of mult-resoluton decomposton of a sgnal x[n] s schematcally shown n Fg. 3. Each stage of ths scheme conssts of two dgtal flters and two down-samplers by 2. The frst flter, h[n] s the dscrete mother wavelet, hgh pass n nature, and the second, g[n] s ts mrror verson, low-pass n nature. The down-sampled outputs of frst hgh-pass and low-pass flters provde the detal, D1 and the approxmaton, A1, respectvely. The frst approxmaton, A1 s further decomposed and ths process s contnued as shown n Fg. 3. The EEG sub bands of a2, d2 and d1are shown n fg. 4. Selecton of sutable wavelet and the number of decomposton levels s very mportant n analyss of sgnals usng the wavelet transformaton. The number of decomposton levels s chosen based on the domnant frequency components of the sgnal. In the present study, snce the EEG sgnals do not have any useful frequency components above 30 Hz, the number of decomposton levels was chosen to be 2. Thus, the EEG sgnals were decomposed nto detals D1 D2 and one fnal approxmaton, A2. Usually, tests are performed wth dfferent types of wavelets and the one, whch gves maxmum effcency, s selected for the partcular applcaton. The smoothng feature of the Daubeches wavelet of order 4 (db4) made t more approprate to detect changes of EEG sgnals. Hence, the wavelet coeffcents were computed usng the db4 n the present study. The proposed method was appled on both data set of EEG data (Sets H and S). In the dscrete wavelet analyss, a sgnal can be represented by ts approxmatons and detals. The detal at level j s defned as D j a j, k j, k ( t ) (1) kz and the approxmaton at level J s defned as A j D j (2) j J It becomes obvous that A and A D j 1 j j (3) f ( t ) A j D j (4) j J Wavelet has several advantages, whch can smultaneously possess compact support, orthogonalty, symmetry, and short support, and hgh order approxmaton. We expermentally found that tme-frequency doman feature provdes superor performance over tme doman feature n the detecton of epleptc EEG sgnals. Fg. 4 Level 2 decomposton of the band-lmted EEG nto three EEG sub bands usng fourth-order Daubeches wavelet (s = a2+d2+d1) Fg. 3 Two level wavelet decomposton
4 C. Feature Extracton The proposed system makes use of a sngle feature called ApEn for the epleptc detecton. The ApEn s a waveletdoman feature that s capable of classfyng complex systems. The value of the ApEn s determned as shown n the followng steps [9], [10]. 1) Let the data sequence contanng N data ponts be X = [x(1), x(2), x(3),..., x(n)]. 2) Let x() be a subsequence of X such that x() = [x(), x( + 1), x( + 2),..., x( + m 1)] for 1 N m, where m represents the number of samples used for the predcton. 3) Let r represent the nose flter level that s defned as r = k SD (5) for k = 0, 0.1, 0.2, 0.3,, 0.9 where SDs the standard devaton of the data sequence X. 4) Let {x(j)} represent a set of subsequences obtaned from x(j) by varyng j from 1 to N. Each sequence x(j) n the set of {x(j)} s compared wth x() and, n ths process, two parameters, namely C m (r) and C m+1 (r) are defned as follows: C m (r) = N m j 1 k N-m (6) where k = 1, f x() x(j) r for 1 j N m 0, otherwse and C m+1 (r) = N m j 1 k N-m (7) wth condtons depcted by (A) as shown at the bottom of the page. 5) We defne Φ m (r) and Φ m+1 (r) as follows: Φ m (r) = N m 1 m ln( C ( r)) N m (8) ApEn features. As t consumes more tme n processng these 60 ApEn features, there s a need to select the best thrty features. Table. 1 shows the extracted features (ApEn) for the sub bands for the sample set A. These best features are selected by our novel approach whch nvolves choosng the feature havng mnmal varance wthn the class and maxmum absolute dfference between the classes. Varance has been calculated for each class of sample set to fnd the mnmal varance. And absolute dfference between classes of sample set to fnd the maxmal dfference. TABLE I FEATURE EXTRACTION SAMPLE DATA SET A Set Sub-bands ApEn D H D2 295 A D S D A E. Probablstc Neural Network Classfer The classfcaton of EEG sgnals nto healthy and epleptc sgnals s done usng the probablstc neural network. The archtecture of the PNN s shown n Fg. 5. In machne learnng, a classfer s essentally a mappng from the feature space to the class space. An Artfcal Neural Network (ANN) mplements such a mappng by usng a group of nterconnected artfcal neurons smulatng the human bran. An ANN can be traned to acheve expected classfcaton results aganst the nput and output nformaton stream, so there may not be a need to provde a specfed classfcaton algorthm. There s no need to tran the network over the entre data set agan, so we use PNN to enable quck updates of our network as more patents data becomes avalable. Our PNN has three layers: the Input Layer, the Radal Bass Layer whch evaluates dstances between the nput vector and rows n the weght matrx, and the Compettve Layer whch determnes the classfcaton wth maxmum probablty of correctness. Dmensons of matrces are marked under ther names. Φ m+1 (r) = N m 1 m1 ln( C ( r)) N m (9) Small values of ApEn mply strong regularty n a data sequence and large values mply substantal fluctuatons [11]. In the proposed approach, ApEn s calculated for one approxmaton and for detaled nformaton such as a2 and d2. D. Feature Selecton As dscussed n the above secton, 30 ApEn features have been obtaned from each sub band leadng to a total of 60
5 Fg. 5 PNN structure, R: number of features, Q: number of tranng samples, K: number of classes. 1) Input Layer: The nput vector, denoted as p, s presented as a black vertcal bar n Fg. 5. The nput layer unt does not perform any computaton and smply dstrbutes the nput to neurons n the pattern layer. On recevng a pattern x from nput layer, the neuron x j of the pattern layer computes ts output usng the below formula. ( x x ) ( x x ) (2 ) 2 T 1 j j j( x ) exp 0.5d d 2 (10) Where d denotes dmenson of the pattern vector x, smoothng parameter and xj s the neuron vector. s the 2) Radal Bass Layer: In the Radal Bass Layer, the vector dstances between nput vector p and the weght vector, made up of each row of the weght matrx W are calculated. Here, the vector dstance s defned as the dot product between two vectors. The dot product between p and the -th row of W produces the -th element of the dstance vector matrx, denoted as W p. The bas vector b s then combned wth W p by an element-by-element multplcaton, represented as n Fg. 5. The result s denoted as n = W p b. The transfer functon n PNN has bult nto a dstance crteron wth respect to a center. In ths paper, we defne t as radbas(n) = e n2. Each element of n s substtuted nto the transfer functon and produces correspondng element of a, the output vector of Radal Bass Layer. We can represent the -th element of a as a = radbas( W p b), where W s the -th row of W, and b s the -th element of bas vector b. Radal Bass Layer Weghts: Each row of W s the feature vector of one tranng sample. The number of rows s equal to the number of tranng samples. Radal Bass Layer Bases: All bases n the radal bass layer are set to ln0.5/s, resultng n radal bass functons that cross 0.5 at weghted nputs of ±s, where s s the spread constant of PNN. Accordng to our experence, s = 0.1 can typcally result n the hghest accuracy. Summaton layer neurons compute the maxmum lkelhood of a pattern x beng classfed nto C, by averagng the output of all neurons that belong to the same class usng P( ) exp (11) ( xx ) ( x xj) N T 1 j x 0.5d d 2 N (2 ) j1 2 3) Compettve Layer: There s no bas n the Compettve Layer. In ths layer, the vector a s frst multpled by the layer weght matrx M, producng an output vector d. The compettve functon C produces a 1 correspondng to the largest element of d, and 0 s elsewhere. The ndex of the 1 s the class of the EEG segment. M s set to a K Q matrx of Q target class vectors. If the -th sample n the tranng set s of class j, then we have a 1 on the j-th row of the -th column of M. The decson layer classfes the pattern x n accordance wth Bayes decson rule based on the output of all summaton layer neurons usng Ĉ( x) arg max P ( x ), 1, 2,.., m (12) Where denotes the estmated class of pattern x, and m s the total number of classes n tranng samples. Hence, PNN employed n ths work possesses 30 nodes n the nput layer and 2 nodes n the output layer (the number of nodes n the output layer s the number of classfcatons of EEG sgnals). The performance of the neural model was evaluated n terms of tranng performance and classfcaton accuraces and the results confrmed that the proposed scheme has potental n classfyng the EEG sgnals. III. RESULTS AND DISCUSSION ApEn values are computed for selected combnatons of m, r, and N. The values of m, r, and N that are used for the experments are as follows: ) m = 1, 2, 3; r = 0% 90% of SD of the data sequence n ncrements of 10%; and N = ApEn values are computed for both normal and epleptc EEG sgnals and are fed as nputs to the two neural networks. Among the avalable 100 EEG data sets, 50 data sets are used for tranng and the remanng data sets are used for testng the performance of the neural networks. The potentalty of the ApEn to dscrmnate the two sgnals, namely, normal and epleptc EEG sgnals depends on the values of m, r, and N. Fg. 6 shows Recever Operatng Characterstcs Curve for the overall detecton accuracy (%) obtaned by the PNN usng ApEn as the nput feature. The expermental results show that our PNN usng wavelet based ApEn can well preserve the most dscrmnant nformaton of EEG sgnals and mprove the performance over the extng system n terms of detecton rate. PNN gves good overall accuracy values n the range 98% - 100%, only for a few combnatons of m, r, and N (e.g., m = 1, r = 0 SD for all the values of N and m = 3, r = 0.2 SD, N = 4097). Though the use of ANNs ncreases the computatonal complexty, the hgh overall detecton accuraces are acheved wth ths system surpasses ts dsadvantage as n any automated sezure detecton system, the detecton of the sezure wth hgh accuracy s of prmary mportance. Where N denotes the total number of samples n class C.
6 ( data.html). REFERENCES Fg. 6. ROC Curve: Overall Classfcaton Rate Our expermental results are based on data sets correspondng to fve dfferent subjects only. The optmum ApEn parameter values obtaned based on ths data may not hold good for a general case. Hence, usng a lnear separator wth known ApEn parameter values may not gve good results n stuatons where a large number of dfferent subjects are nvolved. Ths problem wll not arse n the proposed PNN-based method as t has performed well rrespectve of the ApEn parameter values used. It s shown that our wavelet based ApEn possesses good characterstcs such as robustness n the characterzaton of the epleptc patterns and low computatonal burden. Hence, an automated system usng wavelet based ApEn as the nput feature s best suted for the real tme detecton of the epleptc sezures. The proposed system s based on two types of EEG, namely, EEG sgnals of awake and epleptc subjects. It can be made more robust by acclmatzng t to the other manfestatons of EEG lke sleep EEG. [1] Guler, I., Kymk, M. K., Akn, M., & Alkan, A. AR spectral analyss of EEG sgnals by usng maxmum lkelhood estmaton. Computers n Bology and Medcne, 31, , [2] Zoubr, M., & Boashash, B. Sezure detecton of newborn EEG usng a model approach. IEEE Transactons on Bomedcal Engneerng, 45, , [3] Adel H, Zhou Z, Dadmehr N. Analyss of EEG records n an epleptc patent usng wavelet transform. J Neurosc Methods; 123(1):69 87, [4] Pradhan, N., Sadasvan, P. K., & Arunodaya, G. R. Detecton of sezure actvty n EEG by an artfcal neural network: A prelmnary study. Computers and Bomedcal Research, 29, , [5] Weng, W., & Khorasan, K. An adaptve structure neural network wth applcaton to EEG automatc sezure detecton. Neural Networks, 9, , [6] Petrosan, A., Prokhorov, D., Homan, R., Dashe, R., & Wunsch, D. Recurrent neural network based predcton of epleptc sezures n ntraand extracranal EEG. Neurocomputng, 30, , [7] Folkers, A., Mosch, F., Malna, T., & Hofmann, U. G. Realtme boelectrcal data acquston and processng from 128 channels utlzng the wavelet-transformaton. Neurocomputng, 52 54, , [8] Guler I, Ubeyl ED. Applcaton of adaptve neuro-fuzzy nference system for detecton of electrocardographc changes n patents wth partal eplepsy usng feature extracton. Expert Syst Appl; 27(3):323 30, [9] Subas, A. Automatc recognton of alertness level from EEG by usng neural network and wavelet coeffcents. Expert Systems wth Applcatons, 28, , [10] Kandaswamy, A., Kumar, C. S., Ramanathan, R. P., Jayaraman, S., & Malmurugan, N. Neural classfcaton of lung sounds usng wavelet coeffcents. Computers n Bology and Medcne, 34(6), , [11] M. Boulougoura, E. Wadge, V.S. Kodoganns, H.S. Chowdrey, Intellgent systems for computer-asssted clncal endoscopc mage analyss, 2nd IASTED Int. Conf. on Bomedcal Engneerng, Innsbruck, Austra, pp , 2004 IV. CONCLUSION In ths paper, the neural network namely Probablstc Neural Network (PNN), has been employed for the automated detecton of eplepsy. A robust and computatonally lowntensve feature such as wavelet doman based Approxmate Entropy (ApEn) has been used for the proposed epleptc detecton system and a new approach has been used for feature selecton n order to reduce the dmenson and ncrease the computaton speed. Expermental results show that overall accuraces as hgh as 100% can be acheved by ths system. As the proposed system s based on a sngle feature that has a low computatonal burden, t s best suted for the real-tme detecton of epleptc sezures from ambulatory recordngs. ACKNOWLEDGMENT The authors wsh to thank Andrzejak et al., 2001 for the benchmark eeg dataset whch s publcly avalable:
Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur
Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:
More informationMultigradient for Neural Networks for Equalizers 1
Multgradent for Neural Netorks for Equalzers 1 Chulhee ee, Jnook Go and Heeyoung Km Department of Electrcal and Electronc Engneerng Yonse Unversty 134 Shnchon-Dong, Seodaemun-Ku, Seoul 1-749, Korea ABSTRACT
More informationPop-Click Noise Detection Using Inter-Frame Correlation for Improved Portable Auditory Sensing
Advanced Scence and Technology Letters, pp.164-168 http://dx.do.org/10.14257/astl.2013 Pop-Clc Nose Detecton Usng Inter-Frame Correlaton for Improved Portable Audtory Sensng Dong Yun Lee, Kwang Myung Jeon,
More informationComposite Hypotheses testing
Composte ypotheses testng In many hypothess testng problems there are many possble dstrbutons that can occur under each of the hypotheses. The output of the source s a set of parameters (ponts n a parameter
More informationStatistical machine learning and its application to neonatal seizure detection
19/Oct/2009 Statstcal machne learnng and ts applcaton to neonatal sezure detecton Presented by Andry Temko Department of Electrcal and Electronc Engneerng Page 2 of 42 A. Temko, Statstcal Machne Learnng
More informationVQ widely used in coding speech, image, and video
at Scalar quantzers are specal cases of vector quantzers (VQ): they are constraned to look at one sample at a tme (memoryless) VQ does not have such constrant better RD perfomance expected Source codng
More informationAn Improved multiple fractal algorithm
Advanced Scence and Technology Letters Vol.31 (MulGraB 213), pp.184-188 http://dx.do.org/1.1427/astl.213.31.41 An Improved multple fractal algorthm Yun Ln, Xaochu Xu, Jnfeng Pang College of Informaton
More informationBoostrapaggregating (Bagging)
Boostrapaggregatng (Baggng) An ensemble meta-algorthm desgned to mprove the stablty and accuracy of machne learnng algorthms Can be used n both regresson and classfcaton Reduces varance and helps to avod
More informationA Robust Method for Calculating the Correlation Coefficient
A Robust Method for Calculatng the Correlaton Coeffcent E.B. Nven and C. V. Deutsch Relatonshps between prmary and secondary data are frequently quantfed usng the correlaton coeffcent; however, the tradtonal
More informationEEE 241: Linear Systems
EEE : Lnear Systems Summary #: Backpropagaton BACKPROPAGATION The perceptron rule as well as the Wdrow Hoff learnng were desgned to tran sngle layer networks. They suffer from the same dsadvantage: they
More informationRegularized Discriminant Analysis for Face Recognition
1 Regularzed Dscrmnant Analyss for Face Recognton Itz Pma, Mayer Aladem Department of Electrcal and Computer Engneerng, Ben-Guron Unversty of the Negev P.O.Box 653, Beer-Sheva, 845, Israel. Abstract Ths
More informationCOMPUTATIONALLY EFFICIENT WAVELET AFFINE INVARIANT FUNCTIONS FOR SHAPE RECOGNITION. Erdem Bala, Dept. of Electrical and Computer Engineering,
COMPUTATIONALLY EFFICIENT WAVELET AFFINE INVARIANT FUNCTIONS FOR SHAPE RECOGNITION Erdem Bala, Dept. of Electrcal and Computer Engneerng, Unversty of Delaware, 40 Evans Hall, Newar, DE, 976 A. Ens Cetn,
More informationGeneralized Linear Methods
Generalzed Lnear Methods 1 Introducton In the Ensemble Methods the general dea s that usng a combnaton of several weak learner one could make a better learner. More formally, assume that we have a set
More informationOther NN Models. Reinforcement learning (RL) Probabilistic neural networks
Other NN Models Renforcement learnng (RL) Probablstc neural networks Support vector machne (SVM) Renforcement learnng g( (RL) Basc deas: Supervsed dlearnng: (delta rule, BP) Samples (x, f(x)) to learn
More informationCOMPARING NOISE REMOVAL IN THE WAVELET AND FOURIER DOMAINS
COMPARING NOISE REMOVAL IN THE WAVELET AND FOURIER DOMAINS Robert J. Barsant, and Jordon Glmore Department of Electrcal and Computer Engneerng The Ctadel Charleston, SC, 29407 e-mal: robert.barsant@ctadel.edu
More informationReport on Image warping
Report on Image warpng Xuan Ne, Dec. 20, 2004 Ths document summarzed the algorthms of our mage warpng soluton for further study, and there s a detaled descrpton about the mplementaton of these algorthms.
More informationTracking with Kalman Filter
Trackng wth Kalman Flter Scott T. Acton Vrgna Image and Vdeo Analyss (VIVA), Charles L. Brown Department of Electrcal and Computer Engneerng Department of Bomedcal Engneerng Unversty of Vrgna, Charlottesvlle,
More informationLecture Notes on Linear Regression
Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume
More information2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification
E395 - Pattern Recognton Solutons to Introducton to Pattern Recognton, Chapter : Bayesan pattern classfcaton Preface Ths document s a soluton manual for selected exercses from Introducton to Pattern Recognton
More informationFor now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results.
Neural Networks : Dervaton compled by Alvn Wan from Professor Jtendra Malk s lecture Ths type of computaton s called deep learnng and s the most popular method for many problems, such as computer vson
More informationNegative Binomial Regression
STATGRAPHICS Rev. 9/16/2013 Negatve Bnomal Regresson Summary... 1 Data Input... 3 Statstcal Model... 3 Analyss Summary... 4 Analyss Optons... 7 Plot of Ftted Model... 8 Observed Versus Predcted... 10 Predctons...
More informationComparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method
Appled Mathematcal Scences, Vol. 7, 0, no. 47, 07-0 HIARI Ltd, www.m-hkar.com Comparson of the Populaton Varance Estmators of -Parameter Exponental Dstrbuton Based on Multple Crtera Decson Makng Method
More informationSupport Vector Machines. Vibhav Gogate The University of Texas at dallas
Support Vector Machnes Vbhav Gogate he Unversty of exas at dallas What We have Learned So Far? 1. Decson rees. Naïve Bayes 3. Lnear Regresson 4. Logstc Regresson 5. Perceptron 6. Neural networks 7. K-Nearest
More informationChapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems
Numercal Analyss by Dr. Anta Pal Assstant Professor Department of Mathematcs Natonal Insttute of Technology Durgapur Durgapur-713209 emal: anta.bue@gmal.com 1 . Chapter 5 Soluton of System of Lnear Equatons
More informationAdmin NEURAL NETWORKS. Perceptron learning algorithm. Our Nervous System 10/25/16. Assignment 7. Class 11/22. Schedule for the rest of the semester
0/25/6 Admn Assgnment 7 Class /22 Schedule for the rest of the semester NEURAL NETWORKS Davd Kauchak CS58 Fall 206 Perceptron learnng algorthm Our Nervous System repeat untl convergence (or for some #
More informationarxiv:cs.cv/ Jun 2000
Correlaton over Decomposed Sgnals: A Non-Lnear Approach to Fast and Effectve Sequences Comparson Lucano da Fontoura Costa arxv:cs.cv/0006040 28 Jun 2000 Cybernetc Vson Research Group IFSC Unversty of São
More informationPsychology 282 Lecture #24 Outline Regression Diagnostics: Outliers
Psychology 282 Lecture #24 Outlne Regresson Dagnostcs: Outlers In an earler lecture we studed the statstcal assumptons underlyng the regresson model, ncludng the followng ponts: Formal statement of assumptons.
More informationSolving Nonlinear Differential Equations by a Neural Network Method
Solvng Nonlnear Dfferental Equatons by a Neural Network Method Luce P. Aarts and Peter Van der Veer Delft Unversty of Technology, Faculty of Cvlengneerng and Geoscences, Secton of Cvlengneerng Informatcs,
More informationA Bayes Algorithm for the Multitask Pattern Recognition Problem Direct Approach
A Bayes Algorthm for the Multtask Pattern Recognton Problem Drect Approach Edward Puchala Wroclaw Unversty of Technology, Char of Systems and Computer etworks, Wybrzeze Wyspanskego 7, 50-370 Wroclaw, Poland
More informationLecture 12: Classification
Lecture : Classfcaton g Dscrmnant functons g The optmal Bayes classfer g Quadratc classfers g Eucldean and Mahalanobs metrcs g K Nearest Neghbor Classfers Intellgent Sensor Systems Rcardo Guterrez-Osuna
More informationNumerical Heat and Mass Transfer
Master degree n Mechancal Engneerng Numercal Heat and Mass Transfer 06-Fnte-Dfference Method (One-dmensonal, steady state heat conducton) Fausto Arpno f.arpno@uncas.t Introducton Why we use models and
More informationSee Book Chapter 11 2 nd Edition (Chapter 10 1 st Edition)
Count Data Models See Book Chapter 11 2 nd Edton (Chapter 10 1 st Edton) Count data consst of non-negatve nteger values Examples: number of drver route changes per week, the number of trp departure changes
More informationCSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography
CSc 6974 and ECSE 6966 Math. Tech. for Vson, Graphcs and Robotcs Lecture 21, Aprl 17, 2006 Estmatng A Plane Homography Overvew We contnue wth a dscusson of the major ssues, usng estmaton of plane projectve
More informationEvaluation for sets of classes
Evaluaton for Tet Categorzaton Classfcaton accuracy: usual n ML, the proporton of correct decsons, Not approprate f the populaton rate of the class s low Precson, Recall and F 1 Better measures 21 Evaluaton
More informationMULTISPECTRAL IMAGE CLASSIFICATION USING BACK-PROPAGATION NEURAL NETWORK IN PCA DOMAIN
MULTISPECTRAL IMAGE CLASSIFICATION USING BACK-PROPAGATION NEURAL NETWORK IN PCA DOMAIN S. Chtwong, S. Wtthayapradt, S. Intajag, and F. Cheevasuvt Faculty of Engneerng, Kng Mongkut s Insttute of Technology
More informationDe-noising Method Based on Kernel Adaptive Filtering for Telemetry Vibration Signal of the Vehicle Test Kejun ZENG
6th Internatonal Conference on Mechatroncs, Materals, Botechnology and Envronment (ICMMBE 6) De-nosng Method Based on Kernel Adaptve Flterng for elemetry Vbraton Sgnal of the Vehcle est Kejun ZEG PLA 955
More informationChapter 11: Simple Linear Regression and Correlation
Chapter 11: Smple Lnear Regresson and Correlaton 11-1 Emprcal Models 11-2 Smple Lnear Regresson 11-3 Propertes of the Least Squares Estmators 11-4 Hypothess Test n Smple Lnear Regresson 11-4.1 Use of t-tests
More informationLINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity
LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 30 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 2 Remedes for multcollnearty Varous technques have
More informationP R. Lecture 4. Theory and Applications of Pattern Recognition. Dept. of Electrical and Computer Engineering /
Theory and Applcatons of Pattern Recognton 003, Rob Polkar, Rowan Unversty, Glassboro, NJ Lecture 4 Bayes Classfcaton Rule Dept. of Electrcal and Computer Engneerng 0909.40.0 / 0909.504.04 Theory & Applcatons
More informationChapter 13: Multiple Regression
Chapter 13: Multple Regresson 13.1 Developng the multple-regresson Model The general model can be descrbed as: It smplfes for two ndependent varables: The sample ft parameter b 0, b 1, and b are used to
More informationComparison of Regression Lines
STATGRAPHICS Rev. 9/13/2013 Comparson of Regresson Lnes Summary... 1 Data Input... 3 Analyss Summary... 4 Plot of Ftted Model... 6 Condtonal Sums of Squares... 6 Analyss Optons... 7 Forecasts... 8 Confdence
More informationAppendix B: Resampling Algorithms
407 Appendx B: Resamplng Algorthms A common problem of all partcle flters s the degeneracy of weghts, whch conssts of the unbounded ncrease of the varance of the mportance weghts ω [ ] of the partcles
More informationCHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE
CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE Analytcal soluton s usually not possble when exctaton vares arbtrarly wth tme or f the system s nonlnear. Such problems can be solved by numercal tmesteppng
More informationKernel Methods and SVMs Extension
Kernel Methods and SVMs Extenson The purpose of ths document s to revew materal covered n Machne Learnng 1 Supervsed Learnng regardng support vector machnes (SVMs). Ths document also provdes a general
More informationUnified Subspace Analysis for Face Recognition
Unfed Subspace Analyss for Face Recognton Xaogang Wang and Xaoou Tang Department of Informaton Engneerng The Chnese Unversty of Hong Kong Shatn, Hong Kong {xgwang, xtang}@e.cuhk.edu.hk Abstract PCA, LDA
More informationNon-linear Canonical Correlation Analysis Using a RBF Network
ESANN' proceedngs - European Smposum on Artfcal Neural Networks Bruges (Belgum), 4-6 Aprl, d-sde publ., ISBN -97--, pp. 57-5 Non-lnear Canoncal Correlaton Analss Usng a RBF Network Sukhbnder Kumar, Elane
More information), it produces a response (output function g (x)
Lnear Systems Revew Notes adapted from notes by Mchael Braun Typcally n electrcal engneerng, one s concerned wth functons of tme, such as a voltage waveform System descrpton s therefore defned n the domans
More informationADVANCED MACHINE LEARNING ADVANCED MACHINE LEARNING
1 ADVANCED ACHINE LEARNING ADVANCED ACHINE LEARNING Non-lnear regresson technques 2 ADVANCED ACHINE LEARNING Regresson: Prncple N ap N-dm. nput x to a contnuous output y. Learn a functon of the type: N
More informationSparse Gaussian Processes Using Backward Elimination
Sparse Gaussan Processes Usng Backward Elmnaton Lefeng Bo, Lng Wang, and Lcheng Jao Insttute of Intellgent Informaton Processng and Natonal Key Laboratory for Radar Sgnal Processng, Xdan Unversty, X an
More informationUncertainty in measurements of power and energy on power networks
Uncertanty n measurements of power and energy on power networks E. Manov, N. Kolev Department of Measurement and Instrumentaton, Techncal Unversty Sofa, bul. Klment Ohrdsk No8, bl., 000 Sofa, Bulgara Tel./fax:
More informationSupporting Information
Supportng Informaton The neural network f n Eq. 1 s gven by: f x l = ReLU W atom x l + b atom, 2 where ReLU s the element-wse rectfed lnear unt, 21.e., ReLUx = max0, x, W atom R d d s the weght matrx to
More informationChange Detection: Current State of the Art and Future Directions
Change Detecton: Current State of the Art and Future Drectons Dapeng Olver Wu Electrcal & Computer Engneerng Unversty of Florda http://www.wu.ece.ufl.edu/ Outlne Motvaton & problem statement Change detecton
More informationSDMML HT MSc Problem Sheet 4
SDMML HT 06 - MSc Problem Sheet 4. The recever operatng characterstc ROC curve plots the senstvty aganst the specfcty of a bnary classfer as the threshold for dscrmnaton s vared. Let the data space be
More informationGEMINI GEneric Multimedia INdexIng
GEMINI GEnerc Multmeda INdexIng Last lecture, LSH http://www.mt.edu/~andon/lsh/ Is there another possble soluton? Do we need to perform ANN? 1 GEnerc Multmeda INdexIng dstance measure Sub-pattern Match
More informationLossy Compression. Compromise accuracy of reconstruction for increased compression.
Lossy Compresson Compromse accuracy of reconstructon for ncreased compresson. The reconstructon s usually vsbly ndstngushable from the orgnal mage. Typcally, one can get up to 0:1 compresson wth almost
More informationKernels in Support Vector Machines. Based on lectures of Martin Law, University of Michigan
Kernels n Support Vector Machnes Based on lectures of Martn Law, Unversty of Mchgan Non Lnear separable problems AND OR NOT() The XOR problem cannot be solved wth a perceptron. XOR Per Lug Martell - Systems
More informationCS 3710: Visual Recognition Classification and Detection. Adriana Kovashka Department of Computer Science January 13, 2015
CS 3710: Vsual Recognton Classfcaton and Detecton Adrana Kovashka Department of Computer Scence January 13, 2015 Plan for Today Vsual recognton bascs part 2: Classfcaton and detecton Adrana s research
More information1 Convex Optimization
Convex Optmzaton We wll consder convex optmzaton problems. Namely, mnmzaton problems where the objectve s convex (we assume no constrants for now). Such problems often arse n machne learnng. For example,
More informationPredictive Analytics : QM901.1x Prof U Dinesh Kumar, IIMB. All Rights Reserved, Indian Institute of Management Bangalore
Sesson Outlne Introducton to classfcaton problems and dscrete choce models. Introducton to Logstcs Regresson. Logstc functon and Logt functon. Maxmum Lkelhood Estmator (MLE) for estmaton of LR parameters.
More informationCONTRAST ENHANCEMENT FOR MIMIMUM MEAN BRIGHTNESS ERROR FROM HISTOGRAM PARTITIONING INTRODUCTION
CONTRAST ENHANCEMENT FOR MIMIMUM MEAN BRIGHTNESS ERROR FROM HISTOGRAM PARTITIONING N. Phanthuna 1,2, F. Cheevasuvt 2 and S. Chtwong 2 1 Department of Electrcal Engneerng, Faculty of Engneerng Rajamangala
More informationRBF Neural Network Model Training by Unscented Kalman Filter and Its Application in Mechanical Fault Diagnosis
Appled Mechancs and Materals Submtted: 24-6-2 ISSN: 662-7482, Vols. 62-65, pp 2383-2386 Accepted: 24-6- do:.428/www.scentfc.net/amm.62-65.2383 Onlne: 24-8- 24 rans ech Publcatons, Swtzerland RBF Neural
More informationDigital Signal Processing
Dgtal Sgnal Processng Dscrete-tme System Analyss Manar Mohasen Offce: F8 Emal: manar.subh@ut.ac.r School of IT Engneerng Revew of Precedent Class Contnuous Sgnal The value of the sgnal s avalable over
More informationThe optimal delay of the second test is therefore approximately 210 hours earlier than =2.
THE IEC 61508 FORMULAS 223 The optmal delay of the second test s therefore approxmately 210 hours earler than =2. 8.4 The IEC 61508 Formulas IEC 61508-6 provdes approxmaton formulas for the PF for smple
More informationLogistic Regression. CAP 5610: Machine Learning Instructor: Guo-Jun QI
Logstc Regresson CAP 561: achne Learnng Instructor: Guo-Jun QI Bayes Classfer: A Generatve model odel the posteror dstrbuton P(Y X) Estmate class-condtonal dstrbuton P(X Y) for each Y Estmate pror dstrbuton
More informationDepartment of Electrical & Electronic Engineeing Imperial College London. E4.20 Digital IC Design. Median Filter Project Specification
Desgn Project Specfcaton Medan Flter Department of Electrcal & Electronc Engneeng Imperal College London E4.20 Dgtal IC Desgn Medan Flter Project Specfcaton A medan flter s used to remove nose from a sampled
More informationSEPARATION OF ORIGINAL PAINTINGS OF MATISSE AND HIS FAKES USING WAVELET AND ARTIFICIAL NEURAL NETWORKS
ISTANBUL UNIVERSITY JOURNAL OF ELECTRICAL & ELECTRONICS ENGINEERING YEAR VOLUME NUMBER : 2009 : 9 : 1 (791-796) SEPARATION OF ORIGINAL PAINTINGS OF MATISSE AND HIS FAKES USING WAVELET AND ARTIFICIAL NEURAL
More informationSemi-supervised Classification with Active Query Selection
Sem-supervsed Classfcaton wth Actve Query Selecton Jao Wang and Swe Luo School of Computer and Informaton Technology, Beng Jaotong Unversty, Beng 00044, Chna Wangjao088@63.com Abstract. Labeled samples
More informationDepartment of Computer Science Artificial Intelligence Research Laboratory. Iowa State University MACHINE LEARNING
MACHINE LEANING Vasant Honavar Bonformatcs and Computatonal Bology rogram Center for Computatonal Intellgence, Learnng, & Dscovery Iowa State Unversty honavar@cs.astate.edu www.cs.astate.edu/~honavar/
More informationMAXIMUM A POSTERIORI TRANSDUCTION
MAXIMUM A POSTERIORI TRANSDUCTION LI-WEI WANG, JU-FU FENG School of Mathematcal Scences, Peng Unversty, Bejng, 0087, Chna Center for Informaton Scences, Peng Unversty, Bejng, 0087, Chna E-MIAL: {wanglw,
More informationSequential Condition Diagnosis for Centrifugal Pump System Using Fuzzy Neural Network
eural Informaton Processng Letters and Revews Vol., o. 3, March 007 LETTER Sequental Condton Dagnoss for Centrfugal Pump System Usng Fuzzy eural etwork Huaqng Wang and Peng Chen Department of Envronmental
More informationFUZZY GOAL PROGRAMMING VS ORDINARY FUZZY PROGRAMMING APPROACH FOR MULTI OBJECTIVE PROGRAMMING PROBLEM
Internatonal Conference on Ceramcs, Bkaner, Inda Internatonal Journal of Modern Physcs: Conference Seres Vol. 22 (2013) 757 761 World Scentfc Publshng Company DOI: 10.1142/S2010194513010982 FUZZY GOAL
More informationConvergence of random processes
DS-GA 12 Lecture notes 6 Fall 216 Convergence of random processes 1 Introducton In these notes we study convergence of dscrete random processes. Ths allows to characterze phenomena such as the law of large
More informationCOMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS
Avalable onlne at http://sck.org J. Math. Comput. Sc. 3 (3), No., 6-3 ISSN: 97-537 COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS
More informationEcon107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)
I. Classcal Assumptons Econ7 Appled Econometrcs Topc 3: Classcal Model (Studenmund, Chapter 4) We have defned OLS and studed some algebrac propertes of OLS. In ths topc we wll study statstcal propertes
More informationTutorial 2. COMP4134 Biometrics Authentication. February 9, Jun Xu, Teaching Asistant
Tutoral 2 COMP434 ometrcs uthentcaton Jun Xu, Teachng sstant csjunxu@comp.polyu.edu.hk February 9, 207 Table of Contents Problems Problem : nswer the questons Problem 2: Power law functon Problem 3: Convoluton
More informationCSC 411 / CSC D11 / CSC C11
18 Boostng s a general strategy for learnng classfers by combnng smpler ones. The dea of boostng s to take a weak classfer that s, any classfer that wll do at least slghtly better than chance and use t
More informationBOOTSTRAP METHOD FOR TESTING OF EQUALITY OF SEVERAL MEANS. M. Krishna Reddy, B. Naveen Kumar and Y. Ramu
BOOTSTRAP METHOD FOR TESTING OF EQUALITY OF SEVERAL MEANS M. Krshna Reddy, B. Naveen Kumar and Y. Ramu Department of Statstcs, Osmana Unversty, Hyderabad -500 007, Inda. nanbyrozu@gmal.com, ramu0@gmal.com
More informationA New Evolutionary Computation Based Approach for Learning Bayesian Network
Avalable onlne at www.scencedrect.com Proceda Engneerng 15 (2011) 4026 4030 Advanced n Control Engneerng and Informaton Scence A New Evolutonary Computaton Based Approach for Learnng Bayesan Network Yungang
More informationMultilayer Perceptron (MLP)
Multlayer Perceptron (MLP) Seungjn Cho Department of Computer Scence and Engneerng Pohang Unversty of Scence and Technology 77 Cheongam-ro, Nam-gu, Pohang 37673, Korea seungjn@postech.ac.kr 1 / 20 Outlne
More informationUsing Immune Genetic Algorithm to Optimize BP Neural Network and Its Application Peng-fei LIU1,Qun-tai SHEN1 and Jun ZHI2,*
Advances n Computer Scence Research (ACRS), volume 54 Internatonal Conference on Computer Networks and Communcaton Technology (CNCT206) Usng Immune Genetc Algorthm to Optmze BP Neural Network and Its Applcaton
More informationLecture 10 Support Vector Machines II
Lecture 10 Support Vector Machnes II 22 February 2016 Taylor B. Arnold Yale Statstcs STAT 365/665 1/28 Notes: Problem 3 s posted and due ths upcomng Frday There was an early bug n the fake-test data; fxed
More informationA NEW DISCRETE WAVELET TRANSFORM
A NEW DISCRETE WAVELET TRANSFORM ALEXANDRU ISAR, DORINA ISAR Keywords: Dscrete wavelet, Best energy concentraton, Low SNR sgnals The Dscrete Wavelet Transform (DWT) has two parameters: the mother of wavelets
More informationChapter 7 Channel Capacity and Coding
Wreless Informaton Transmsson System Lab. Chapter 7 Channel Capacty and Codng Insttute of Communcatons Engneerng atonal Sun Yat-sen Unversty Contents 7. Channel models and channel capacty 7.. Channel models
More informationRolling Bearing Fault Diagnosis Based on EMD-TEO and Mahalanobis Distance
Rollng Bearng Fault Dagnoss Based on EMD-TE and Mahalanobs Dstance Lu Chen, Hu Jameng, Lu Hongme and Wang Jng RLLG BEARG FAULT DAGSS BASED EMD-TE AD MAHALABS DSTACE. LU CHE, HU JAMEG, LU HGME AD WAG JG
More informationCLASSIFICATION OF EEG SIGNALS USING ADAPTIVE TIME-FREQUENCY DISTRIBUTIONS
Metrol. Meas. Syst., Vol. 3 (06), No., pp. 5 60. MEROLOGY AND MEASUREMEN SYSEMS Index 330930, ISSN 0860-89 www.metrology.pg.gda.pl CLASSIFICAION OF EEG SIGNALS USING ADAPIVE IME-FREQUENCY DISRIBUIONS Nabeel
More informationInductance Calculation for Conductors of Arbitrary Shape
CRYO/02/028 Aprl 5, 2002 Inductance Calculaton for Conductors of Arbtrary Shape L. Bottura Dstrbuton: Internal Summary In ths note we descrbe a method for the numercal calculaton of nductances among conductors
More informationWhite Noise Reduction of Audio Signal using Wavelets Transform with Modified Universal Threshold
Whte Nose Reducton of Audo Sgnal usng Wavelets Transform wth Modfed Unversal Threshold MATKO SARIC, LUKI BILICIC, HRVOJE DUJMIC Unversty of Splt R.Boskovca b.b, HR 1000 Splt CROATIA Abstract: - Ths paper
More informationMore metrics on cartesian products
More metrcs on cartesan products If (X, d ) are metrc spaces for 1 n, then n Secton II4 of the lecture notes we defned three metrcs on X whose underlyng topologes are the product topology The purpose of
More informationLinear Feature Engineering 11
Lnear Feature Engneerng 11 2 Least-Squares 2.1 Smple least-squares Consder the followng dataset. We have a bunch of nputs x and correspondng outputs y. The partcular values n ths dataset are x y 0.23 0.19
More informationTransfer Functions. Convenient representation of a linear, dynamic model. A transfer function (TF) relates one input and one output: ( ) system
Transfer Functons Convenent representaton of a lnear, dynamc model. A transfer functon (TF) relates one nput and one output: x t X s y t system Y s The followng termnology s used: x y nput output forcng
More informationDiscretization of Continuous Attributes in Rough Set Theory and Its Application*
Dscretzaton of Contnuous Attrbutes n Rough Set Theory and Its Applcaton* Gexang Zhang 1,2, Lazhao Hu 1, and Wedong Jn 2 1 Natonal EW Laboratory, Chengdu 610036 Schuan, Chna dylan7237@sna.com 2 School of
More informationNatural Images, Gaussian Mixtures and Dead Leaves Supplementary Material
Natural Images, Gaussan Mxtures and Dead Leaves Supplementary Materal Danel Zoran Interdscplnary Center for Neural Computaton Hebrew Unversty of Jerusalem Israel http://www.cs.huj.ac.l/ danez Yar Wess
More informationNeural Networks & Learning
Neural Netorks & Learnng. Introducton The basc prelmnares nvolved n the Artfcal Neural Netorks (ANN) are descrbed n secton. An Artfcal Neural Netorks (ANN) s an nformaton-processng paradgm that nspred
More information4 Analysis of Variance (ANOVA) 5 ANOVA. 5.1 Introduction. 5.2 Fixed Effects ANOVA
4 Analyss of Varance (ANOVA) 5 ANOVA 51 Introducton ANOVA ANOVA s a way to estmate and test the means of multple populatons We wll start wth one-way ANOVA If the populatons ncluded n the study are selected
More informationNumber of cases Number of factors Number of covariates Number of levels of factor i. Value of the dependent variable for case k
ANOVA Model and Matrx Computatons Notaton The followng notaton s used throughout ths chapter unless otherwse stated: N F CN Y Z j w W Number of cases Number of factors Number of covarates Number of levels
More informationFeb 14: Spatial analysis of data fields
Feb 4: Spatal analyss of data felds Mappng rregularly sampled data onto a regular grd Many analyss technques for geophyscal data requre the data be located at regular ntervals n space and/or tme. hs s
More informationErrors for Linear Systems
Errors for Lnear Systems When we solve a lnear system Ax b we often do not know A and b exactly, but have only approxmatons  and ˆb avalable. Then the best thng we can do s to solve ˆx ˆb exactly whch
More informationx = , so that calculated
Stat 4, secton Sngle Factor ANOVA notes by Tm Plachowsk n chapter 8 we conducted hypothess tests n whch we compared a sngle sample s mean or proporton to some hypotheszed value Chapter 9 expanded ths to
More informationLinear Approximation with Regularization and Moving Least Squares
Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...
More informationDr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur
Analyss of Varance and Desgn of Exerments-I MODULE III LECTURE - 2 EXPERIMENTAL DESIGN MODELS Dr. Shalabh Deartment of Mathematcs and Statstcs Indan Insttute of Technology Kanur 2 We consder the models
More information