A General Method for Assessing the Uncertainty in Classified Remotely Sensed Data at Pixel Scale

Size: px
Start display at page:

Download "A General Method for Assessing the Uncertainty in Classified Remotely Sensed Data at Pixel Scale"

Transcription

1 Proceedngs of the 8th Internatonal Symposum on Spatal Accuracy Assessment n Natural Resources and Envronmental Scences Shangha, P. R. Chna, June 25-27, 2008, pp A General ethod for Assessng the Uncertanty n Classfed Remotely Sensed Data at Pxel Scale Yanchen Bo, 2 + and Jnfeng Wang 3. State Key Laboratory of Remote Sensng Scence, Jontly Sponsored by Bejng Normal Unversty and Insttute of Remote Sensng Applcatons, CAS, 2. Research Center for Remote Sensng and GIS, School of Geography, Bejng Key Laboratory for Remote Sensng of Envronment and Dgtal Ctes, Bejng Normal Unversty, Bejng, 00875, Chna 3. LREIS, Insttute of Geographcal Scence and Natural Resources Research Chnese Academy of Scences, Bejng, 000, Chna Abstract. The uncertanty assessment on the classfcaton of remotely sensed data s a crtcal problem n both academc arena and applcatons. The conventonal soluton to ths problem s based on the error matrx (.e. confuson matrx) and the kappa statstcs derved from the error matrx. However, no spatal dstrbuton nformaton of classfcaton uncertanty can be presented by ths method. A probablty vector-based method has been developed for assessng classfcaton at pxel-scale. However, the use of ths method s severely lmted because the probablty vector can be derved only through the Bayesan classfcaton. In practce, other classfers such as Artfcal Neural Network Classfer, nmum Dstance Classfer, ahalanobs Dstance Classfer and Fuzzy Classfer are wldly used for remote sensng data classfcaton. To assess the uncertanty of thematc maps by these classfers at pxel-scale, a general method s presented n ths paper whch extends the probablty vector-based method to the assessment on the uncertanty classfed by classfers beyond Bayesan classfer. Ths extenson s realzed through a transformaton method, whch transforms the embershp Vector n varous classfers to the transformed probablty vector so that t s comparable to the probablty vector n Bayesan Classfer. The uncertanty measurements could be derved from the probablty vector were evaluated and, the probablty resdual and entropy that derved from the extended probablty vector are used as ndcators to assess the absolute and relatve uncertanty perceptvely. The uncertantes by dfferent classfers are compared at pxel scale. Some examples of the uncertanty of maps from dstance classfers were presented and compared wth that of maps from LC classfer. Keywords: scale, a posteror probablty vector, uncertanty measure.. Introducton Categorcal spatal data set such as land use/land cover and vegetaton type derved from remotely sensed data s the crtcal nputs n ecologcal and envronmental modelng. As the classfed maps from remotely sensed data are uncertan n nature, assessment on the accuracy and uncertanty of remotely sensed data classfcaton s essental to the propagaton and assessment on the uncertanty of models output and, further more, the rsks of decsons based on the models outputs. An essental aspect of the ncreasng sophstcaton of ecologcal and envronmental models s the use of spatally explct nputs and outputs. Thus, the spatal dstrbuton of the classfed maps from remotely sensed data has become a new challenge. The wdely accepted method for assessng the accuracy of thematc maps from remotely sensed data has been the error matrx, or confuson matrx (Congalton and Green, 999). A lmtaton of the error matrx s that t allows only one reference class for each reference ste. Ths s problematc n stes where selecton of + Correspondng author. Tel.: ; fax: E-mal address: boyc@bnu.edu.cn ISBN: , ISBN3:

2 only one class s dffcult or napproprate. A soluton to ths problem s developed by usng fuzzy set theory (Gopal and Woodcock, 994; Woodcock and Gopal, 2000). A common drawback of error matrx based method and fuzzy set theory based method s that t can not provde the spatal dstrbuton of classfcaton uncertanty. Assessng on the uncertanty of classfcaton based on the LC a posteror probablty vector was proposed by Goodchld et al. (992). several uncertanty measures were derved from the LC a posteror probablty vector by Sh (998). However, the practcablty of per-pxel assessment on classfcaton uncertanty based on the a posteror probablty vector was lmtng because the a posteror probablty vector can only be derved through LC classfcaton. For other wdely used classfers n remotely sensed data classfcaton, such as nmum Dstance Classfer ahalabous Dstance Classfer, Artfcal Neural Network Classfer and Fuzzy Classfer, how to assess the uncertanty of classfcaton at pxel-scale s reman problematc. In addton, the varous uncertanty measures were developed based on the LC a posteror probablty vector. Evaluatng and selectng approprate uncertanty measures for uncertanty assessment on remotely sensed data classfcaton at pxel-scale s mportant to the uncertanty assessment practces. The research objectves of ths paper are twofold. Frstly, select the approprate measures for uncertanty assessment on remotely sensed data classfcaton at pxel scale. Secondly, develop a general method for assessng the uncertanty of classfcaton at pxel scale so that the uncertanty measures derved LC a posteror probablty vector can be used to the assessment on the uncertanty of maps classfed by classfers beyond LC at pxel scale. In secton 2, the LC a posteror probablty vector was ntroduced, the varous uncertanty measures derved from a posteror probablty vector were evaluated and approprate measures were selected. In secton 3, the extended probablty vector method were developed to extend the uncertanty measures at pxel scale from the assessment on uncertanty of LC classfcaton to the assessment on the uncertanty of ANN classfcaton, Dstance classfcaton and Fuzzy classfcaton. Secton 4 gves some examples of uncertanty assessment at pxel usng measures from a posteror probablty vector on the maps classfed by classfers beyond LC. Secton 5 summarzes the whole paper and dscussons on further studes were gven. 2. The a Posteror Probablty Vector and the Uncertanty easures 2.. The a Posteror Probablty Vector In the LC classfcaton, the decson of assgnng a pxel to a specfc class s based on the a posteror probablty vector whch presents the condtonal probablty the pxel belongs to every pre-defned class n the classfcaton scheme. ost often, the pxel was assgned to the class wth the maxmum probablty value n the probablty vector. Let the spectral classes for an mage be represented by ω, =... () where s the total number of classes. Accordng to Rchard et al. (998), to determne the class to whch a pxel at a locaton x belongs, the condtonal probabltes are of nterest. p( ω, =,... (2) where the poston vector x s a column vector of brghtness values for the pxel,t descrbes the pxel as a pont n multspectral space wth coordnates defned by the brghtness. The probablty p( ω gves the lkelhood that the correct class sω for a pxel at poston x. Classfcaton s performed accordng to x ω f p( ω > p( ω for all j (3) j.e., the pxel at x belongs to classω f p( ω s the largest. 87

3 Often, the p( ω n () s unknown. Suppose that suffcent tranng data s avalable for each ground cover type, the probablty dstrbuton for a cover type can be estmated. The desred p( ω and the pxω ( ) estmated from tranng data are related by Bayes theorem: p( ω p( x ω ) p( ω )/ p( x) = (4) where p( ω ) s the probablty that classω occurs n the mage. p(x) s the probablty of fndng a pxel from any class at locaton x and can be expressed as: px ( ) = px ( ω) p( ω) = (5) Where the p( ω ) are called a pror probablty and p( ω are a posteror probablty. Usng (2), the classfcaton rule of () s: x ω, f p( x ω ) p( ω ) > p( x ω ) p( ω ) for all j (6) j j where the p( x) has been removed as a common factor. For mathematcal convenence, defne: then (6) s reformed as: g ( x) = ln{ p( x ω ) p( ω )} = ln p( x ω ) + ln p( ω ) (7) x ω, f g ( x) > g ( x) for all j (8) j Ths s the decson rule used n maxmum lkelhood classfcaton. pxω ( ) are referred to as dscrmnate functons. Assume that the probablty dstrbutons for the classes are of multvarate normal dstrbuton, then the dscrmnant functon can be restated as: g x = p x m x m (9) T ( ) ln ( ω) ln ( ) ( ) where m and are the mean vector and covarance matrx of the data n class ω. They were estmated from tranng data. equaton(3)mples that whle the pxel at poston x was assgned to class ω, t was not 00%sure that pxel at x belongs to class ω. Thus, ths s an uncertan decson. The posteror probablty n the dscrmnate functon s an ndcator of the uncertanty n classfcaton decson. In equaton(4), the essence n classfcaton decson s to estmate the condtonal probablty at whch the pxel at x belongs to class ω, =,.... The posteror condtonal probablty pxel at x belongs to every class composed a posteror probablty vector: [ p( ω, p( ω,..., p( ω,..., p( ω ] T (0) 2 Remove the elements of zero value n the posteror probablty vector and rank the elements from large to small, a ranked posteror probablty vector can be formed (Sh, 998): 88 PV ( x) [ p( ω, p( ω,..., p( ω,..., p( ω ] T = () l l2 l lk where k s the number of non-zero elements n equaton (0). In equaton(), p( ω p( ω f < j Uncertanty measures derved from a posteror probablty vector Varous uncertanty measures are derved from posteror probablty vector. Sh (998) defned 4 measures based on the ranked posteror probablty vector. Absolute Uncertanty l lj

4 Absolute Uncertanty was defned as: U ( x) = p( ω /[ p( ω / x)] (2) A l l n equaton(2), U A ranges n a nterval (0, + ). The larger the U A,the more certan the class ω l was assgned to the pxel at x. Relatve Uncertanty Relatve uncertanty descrbes the msclassfcaton of pxel at x between class ω l and classω lj. The relatve uncertanty was defned as: U ( x,, j) = p( ω / x) p( ω / x) (3) R l lj U (,, ) R x j n equaton (3) ranges n a nterval of [0,].The larger the value of UR ( x,, j ),the easer the pxel at x to be dscrmnated between class ω and ω. xture degree of pxel xture degree of pxel measures the degree that a pxel s a mxed pxel. It s defned as: l lj n j= px ( ω ) px ( ω ) = for all j px ( ω ) j (4) Where px ( ω ) s the probablty that pxel at x belongs to classω. The smlar the elements n the posteror probablty vector, the smaller the value, and the more possblty that the pxel x s a mxed pxel. Evdence Incompleteness In the process of remotely sensed data classfcaton, some pxels are hard to determne ther classes and were labeled as unclassfed because the evdence for classfcaton s ncomplete. The evdence ncompleteness ( Θ ) was defned as: k p( ωl + Θ= (5) = Where the value of Θ range n the nterval of [0,]. Large value of evdence ncompleteness ndcates hgh possblty that the pxel was labeled as unclassfed. Relatve axmum Probablty Devaton(Clark Lab, 200) There s another uncertanty measure developed by Clark Lab (200) that we call the Relatve axmum Probablty Devaton. The Relatve axmum Probablty Devaton was defned as: U sum max = (6) Where max stand for the maxmum value n probablty vector. Sum stands for the sum of all elements n probablty. s the number of elements n probablty vector. Probablty Entropy In addton to the uncertanty measures defned by Sh (998), probablty entropy derved from posteror probablty entropy was wldly used as a uncertanty measures of remotely sensed data classfcaton ( asell, 994; Foody, 996 ). Probablty entropy s a measure of uncertanty and nformaton formulated n terms of probablty theory, whch expresses the relatve support assocated wth mutually exclusve alternatve classes. When two or more alternatve classes have non-zero probabltes 89

5 assocated wth them then each probablty s n conflct wth the others. When there s a fnte set of alternatve classes the expected value of conflct s gven by the probablty entropy (Foody, 996). Ths may be used to descrbe the varatons n class membershp probabltes assocated wth each pxel. The probablty entropy ndcates further requred nformaton content to assgn a pxel to a class wth 00% confdence. The probablty entropy was expressed as: = (7) H ( p ) p ( ω x )*log p ( ω x ) =,... 2 For the convenence of vsual presentaton, the relatve probablty entropy was developed by asell et al.(994)and was used n practce. The relatve probablty entropy was defned as: H ( p) H( p)/ H ( p)*00 R = (8) max 3. Selectng Uncertanty easures and Extendng Posteror Probablty Vector 3.. Selectng measures for uncertanty assessment at pxel scale To make the assessment on the uncertanty of remotely sensed data classfcaton at pxel scale be comprehensve and comparable, approprate uncertanty measures have to be selected. The absolute uncertanty defned by Sh (998) and the Relatve axmum Probablty Devaton present smlar uncertanty nformaton. In the defntons of these two measures, the element wth maxmum value n the probablty vector s emphaszed. These two measures are smple and straghtforward n descrbng the absolute uncertanty n assgnng a pxel to a specfc class. The frst drawback of these two measures s that the relatonshps between elements n probablty vector were not consdered. ost often, relatonshp between elements n probablty vector, especally the relatonshp between the maxmum one and the second maxmum one, provde much more uncertanty nformaton. Take a classfcaton wth 3 classes as an example, two posteror probablty vectors [0.5, 0.3, 0.2] and [0.5, 0.4, 0.] present qute dfferent uncertanty nformaton. Though the absolute uncertanty and the relatve maxmum probablty devaton of these two probablty vector are the same, the former mples less uncertanty then the later one. Another drawback of these two uncertanty measures s that values range n a unlmted nterval between (0, + ), whch make t hard to be presented by a gray scale mage. The relatve uncertanty compares the possble msclassfcaton of pxel between every two classes. In relatve uncertanty, relatonshp between elements n probablty vector s consdered. However, because the dfferences of posteror probablty between every two classes have to be calculated, the relatve uncertanty was presented by many posteror probablty dfferences together. Take a classfcaton wth 5 classes as an example, 0 relatve uncertanty values need to be calculated, and for a classfcaton wth 0 classes, 45 relatve probablty values have to be calculated! The mxture degree of pxel () descrbes the degree to whch the pxel beng classfed s a mxed classfcaton. Whle the dfferences between elements n probablty vector are large enough, the value s senstve to the mxture degree of pxel. However, f the maxmum of elements n two probablty vector are the same, s not senstve. Take the two probablty vector [0.6, 0.4, 0.0] and [0.6, 0.2, 0.2] agan as example, both values of these two probablty vector are.33. The evdence ncompleteness s subject to the posteror probablty threshold set n the LC classfcaton. In the LC classfcaton, a probablty threshold was set. If the maxmum value of the probablty vector of a pxel s less than the threshold, then there s no suffcent evdence to assgn ths pxel to any classes. Thus, dfferent posteror probablty threshold comes up wth dfferent mxture degree for same pxel. The probablty entropy was wldly used as an uncertanty measure. As all the elements n the probablty vector were accounted n the calculaton of probablty entropy, all the nformaton n the probablty vector was fully exploted. Because probablty entropy mples the varaton and dsperson of the elements n 90

6 probablty vector (asell, 994), both the nformaton expressed by relatve uncertanty and the mxture degree of pxel were contaned n the probablty entropy. An mportant mert of probablty entropy s that t expresses the uncertanty contaned n probablty vector usng a sngle value. Though the probablty entropy s not senstve to the class that pxel was fnally assgned(zhu, 997), because the pxels to be classfed are always assgned to the class wth maxmum probablty value n the probablty vector, thus the absolute uncertanty and the relatve uncertanty nformaton were both mpled n the probablty entropy. The probablty entropy has been accepted as a good measure of classfcaton uncertanty(zhu,997; Foody, 992; asell, 994). In summary, the absolute uncertanty and maxmum relatve probablty devaton both ndcate the absolute uncertanty nformaton contaned n probablty vector, but no relatve uncertanty nformaton mpled because the relatonshps between elements n probablty vector were not accounted n. Take the uncertanty measures developed by Sh (998) as a whole, they present both absolute and relatve uncertanty nformaton contaned n probablty vector, however need many measures smultaneously. Probablty entropy expresses the uncertanty nformaton contaned n the probablty vector usng a sngle value. Because the relatonshps between elements n probablty vector were accounted n probablty entropy, the relatve uncertanty nformaton contaned n probablty vector was mpled n probablty entropy. Thus, we use the relatve maxmum probablty devaton and the probablty entropy to present the absolute and relatve uncertanty contaned n probablty vector respectvely Extendng the a posteror probablty vector As was dscussed n secton 2, the posteror probablty vector could only be derved through LC classfcaton. In practce, many others classfers ncludng ANN classfer, Fuzzy classfer and varous dstance classfers were wdely used n remotely sensed data classfcaton. In these classfers, there are also measures that present the membershps pxel belongs to every class (Foody, 996; 2000), and can be taken as an approxmaton of posteror probablty that pxel belongs to gven class. Another type of classfer wldly used n remotely sensed data classfcaton s spectral dstance based classfers, ncludng nmum Dstance Classfer and ahalanobs Dstance Classfer (Rchards and Ja, 998). Lke n the LC classfcaton where the membershps of a pxel to every class were presented by the posteror probabltes of the pxel belong to every class, the membershps of a pxel to every class n dstance classfers were presented by the Eucldean dstances or ahalanobs dstances from the feature of pxel to the features of classes estmated from tranng data. Thus, the membershps can also compose a Dstance Vector [ d( ω, d( ω,..., d( ω,..., d( ω ] T (9) 2 In the classfcaton decson of dstance classfers, pxels were assgned to the class wth the mnmum dstance n the dstance vector. Lke the probablty vector, the dstance vector contans the uncertanty nformaton of the dstance-based classfcaton. However, because the value of elements n dstance vector ranges n a unlmted nterval, rather than n the nterval of [0,] n probablty vector, dstance vectors from dfferent classfcatons are not comparable, and the measures of uncertanty dscussed n secton 2 can not derved drectly from the dstance vector. Take three classes classfcaton as an example. Gven the means vector estmated from the tranng data s[30,60, 90], the spectral DN value of pxel A s 70, and pxel B 00. Based on the mnmum dstance classfcaton, the dstance vector of pxel A and pxel B are [40, 0, 20] and [70,50, 0] respectvely. Clearly t s unreasonable to draw a concluson that ths two dstance vectors present same absolute uncertanty just because they have same mnmum element. Thus, t s necessary to convert the elements n the dstance vector to the nterval of [0,] whle the relatve relatonshp between elements are preserved. The conversaton was made usng the followng equaton: 9

7 / d( ω pc( ω = / d( ω (20) = Applyng the equaton (20) to the example above, we can see that the conversed probablty vector for pxel A s [0.57, 0.43, 0.286], and pxel B the [0.534, 0.333, 0.333]. From these two probablty vector, t s reasonable to draw a concluson that the absolute uncertanty of pxel A s less than that of pxel B. By the converson, uncertanty measures from posteror probablty vector n LC can be extended to the assessment on uncertanty of classfcaton from neural network classfer and dstance classfers at pxel scale. Thus, we call the actvaton level vector and the conversed dstance vector together the Extended probablty vector. 4. Example The land cover mappng by multspectral remotely sensed data classfcaton was used here to llustrate the uncertanty assessment at pxel scale presented n ths paper. The standard false color composed magery of Landsat Thematc apper n Laner Lake, U.S. n Fgure. Land cover types n the magery nclude agrcultural feld, bare ground, grass land, confers, decduous, water, urban and the developed. Because there are some cloud and the shade of cloud n ths mage, the classes n classfcaton nclude the cloud and the shade the magery was classfed by LC classfer and numum Dstance classfer respectvely usng same tranng samples. The classfed land cover maps were shown n fgure. The posteror probablty vector was saved n the LC classfcaton. The dstance vector was saved n the mnmum dstance classfcaton and was conversed to the probablty vector usng equaton (28). The axmum Relatve Probablty Devaton and the Probablty Entropy for both LC classfcaton and nmum Dstance classfcaton were calculated and presented n Fgure 2. Fg.. The Landsat T magery and the land cover maps classfed usng axmum Lkelhood Classfer and nmum Dstance Classfer: (a) The Landsat T magery n Laner Lake, U.S.A (b) Land cover map classfed usng LC (c) Land cover map classfed usng nmum Dstance Classfer (d) Land cover legend 92

8 The absolute uncertanty of landcover map classfed usng LC was presented usng the maxmum relatve probablty devaton n Fgure 2 (a). The pxels wth uncertanty larger than 0.5 manly locate n the urban and the developed, bare ground, grass and agrculture feld and the boundary of water. These spatal dstrbuton of absolute uncertanty was caused by the spectral ambguty between urban and the bare ground, and between grass land and agrculture feld. Pxels wth uncertanty between [0.3, 0.5] locates n the confers and the decduous. Ths spatal dstrbuton was caused by the mxed pxels of the confers and the decduous. The absolute uncertanty decreases from the objects center to ther boundares. The relatve uncertanty of land cover map classfed usng LC was presented usng probablty entropy n Fgure 2 (b), n whch the pxels wth hgh probablty entropy locate n the boundary of the water, the clouds, the shades, the confers and the decduous. The uncertantes of these pxels were manly caused by pxel mxture. Fg. 2. Uncertanty of maps classfed usng LC and nmum Dstance Classfer: (a) The axmum Relatve Probablty Devaton of land cover map classfed usng LC; (b) Probablty entropy of land cover map classfed usng LC; (c) axmum relatve probablty devaton of land cover map classfed usng nmum Dstance classfer; (d) Probablty entropy of land cover map classfed usng mnmum dstance classfer The absolute uncertanty and relatve uncertanty of land cover map classfed usng the mnmum dstance classfer were presented n Fgure 2 (c) and fgure 2 (d) respectvely. The maxmum relatve probablty devaton and the probablty entropy were derved from the probablty vector converted from the dstance vector produced by the mnmum dstance classfer. Due to the characterstcs of the mnmum dstance classfer, the classes wth smlar spectral characterstcs, such as urban areas, clouds and bare ground, showed hgh uncertanty. The water showed less uncertanty as ts spectral characterstcs has dstnctve dfference wth other classes. Fgure 2 showed that the uncertanty of maps classfed usng dfferent classfers are qute dfferent even usng same tranng samples. 5. Conclusons The uncertanty assessment on the classfcaton of remotely sensed data s a crtcal problem n both academc arena and applcatons. The conventonal soluton to ths problem s based on the error matrx (.e. confuson matrx) and the kappa statstcs derved from the error matrx. However, no spatal dstrbuton nformaton of classfcaton uncertanty can be presented by ths method. A probablty vector-based method 93

9 has been developed for assessng classfcaton at pxel-scale. However, the use of ths method s severely lmted because the probablty vector can be derved only through the maxmum lkelhood classfcaton and because lack of the comprehensve and comparable uncertanty measures derved from the posteror probablty. The researches n ths paper focused on the solutons to overcome these two drawbacks. Frstly, the uncertanty measures from the probablty vector were evaluated. The maxmum relatve probablty devaton and the probablty entropy were selected to present the absolute uncertanty and relatve uncertanty respectvely. Secondly, the actvaton level vector produced n the feedforward neural network wth backpropagaton algorthm classfcaton and the dstance vector produced n the dstance classfers were converted to probablty vector so that all elements n t are n an nterval of [0,] and are comparable to the probablty vector produced n the L classfer. By ths converson, uncertanty measures for uncertanty assessment on remotely sensed data classfcaton usng LC at pxel scale can be used n the uncertanty assessment on classfcaton usng neural network classfer and dstance classfers at pxel scale. The examples of land cover mappng by classfyng Landsat T multspectral remotely sensed data showed that the maxmum relatve probablty devaton and the probablty entropy can present the spatal dstrbuton of absolute uncertanty and relatve uncertanty well. The spatal dstrbutons of uncertanty of maps classfed usng dfferent classfers are qute dfferent. 6. Acknowledgements Ths research were supported by the grants from Natonal Basc Research Program of Chna (Project No.2007CB ), State Key Lab. of Remote Sensng Scence, NSFC ( ),and the Program for Changjang Scholars and Innovatve Research Team n Unversty(PCSIRT, No.IRT0409) 7. References [] Clark Lab, Idrs Gude to GIS and mage processng, 200, 2. [2] R.G. Congalton and K. Green, Assessng the accuracy of remotely sensed data: Prncples and practces. Lews Publshers, 999: 37. [3] G.. Foody, N.A. Campbell, et al, Dervaton and applcatons of probablstc measures of class membershp from the maxmum lkelhood classfcaton. Photogrammetrc Engneerng and Remote Sensng, 992: [4] G.. Foody, Approaches for the producton and evaluaton of fuzzy land cover classfcatons from remotely sensed data. Internatonal Journal of Remote Sensng, 996: [5] G.. Foody, appng land cover from remotely sensed data wth a softened feed forward neural network classfcaton. Journal of Intellgent and Robotc Systems, 2000: [6] S. Gopal, and C. E. Woodcock, Accuracy Assessment of Thematc aps Usng Fuzzy Sets I: Theory and ethods. Photogrammetrc Engneerng and Remote Sensng, 994: [7] F. asell, C. Conese and L. Petkov, Use of probablty entropy for the estmaton and graphcal representaton of the accuracy of maxmum lkelhood classfcatons. ISPRS Journal of Photogrammetry and Remote Sensng, 994: [8] Rchards and Ja, Remote Sensng dgtal mage analyss: an ntroducton. Sprnger, 998. [9] Wenzhong Sh, Theory and ethods for Error Processng n Spatal Data. Scence Press, 998. [0] C.E. Woodcock, and S. Gopal, Fuzzy set theory and thematc maps: accuracy assessment and area estmaton. Internatonal Journal of Geographcal Informaton Scence, 2000: [] A.X. Zhu, easurng uncertanty n class assgnment for natural resource maps under fuzzy logc. Photogrammetrc engneerng and Remote Sensng997:

International Journal of Mathematical Archive-3(3), 2012, Page: Available online through ISSN

International Journal of Mathematical Archive-3(3), 2012, Page: Available online through   ISSN Internatonal Journal of Mathematcal Archve-3(3), 2012, Page: 1136-1140 Avalable onlne through www.ma.nfo ISSN 2229 5046 ARITHMETIC OPERATIONS OF FOCAL ELEMENTS AND THEIR CORRESPONDING BASIC PROBABILITY

More information

Fuzzy Boundaries of Sample Selection Model

Fuzzy Boundaries of Sample Selection Model Proceedngs of the 9th WSES Internatonal Conference on ppled Mathematcs, Istanbul, Turkey, May 7-9, 006 (pp309-34) Fuzzy Boundares of Sample Selecton Model L. MUHMD SFIIH, NTON BDULBSH KMIL, M. T. BU OSMN

More information

Lecture 12: Classification

Lecture 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 information

EEE 241: Linear Systems

EEE 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 information

MULTISPECTRAL IMAGE CLASSIFICATION USING BACK-PROPAGATION NEURAL NETWORK IN PCA DOMAIN

MULTISPECTRAL 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 information

Uncertainty in measurements of power and energy on power networks

Uncertainty 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 information

A Bayes Algorithm for the Multitask Pattern Recognition Problem Direct Approach

A 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 information

Analysis of Maximum Likelihood Classification. on Multispectral Data

Analysis of Maximum Likelihood Classification. on Multispectral Data Appled Mathematcal Scences, Vol. 6, 0, no. 9, 645-6436 Analyss of Maxmum Lkelhood Classfcaton on Multspectral Data Asmala Ahmad Department of Industral Computng Faculty of Informaton and Communcaton Technology

More information

P R. Lecture 4. Theory and Applications of Pattern Recognition. Dept. of Electrical and Computer Engineering /

P 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 information

On an Extension of Stochastic Approximation EM Algorithm for Incomplete Data Problems. Vahid Tadayon 1

On an Extension of Stochastic Approximation EM Algorithm for Incomplete Data Problems. Vahid Tadayon 1 On an Extenson of Stochastc Approxmaton EM Algorthm for Incomplete Data Problems Vahd Tadayon Abstract: The Stochastc Approxmaton EM (SAEM algorthm, a varant stochastc approxmaton of EM, s a versatle tool

More information

DETERMINATION OF UNCERTAINTY ASSOCIATED WITH QUANTIZATION ERRORS USING THE BAYESIAN APPROACH

DETERMINATION OF UNCERTAINTY ASSOCIATED WITH QUANTIZATION ERRORS USING THE BAYESIAN APPROACH Proceedngs, XVII IMEKO World Congress, June 7, 3, Dubrovn, Croata Proceedngs, XVII IMEKO World Congress, June 7, 3, Dubrovn, Croata TC XVII IMEKO World Congress Metrology n the 3rd Mllennum June 7, 3,

More information

Power law and dimension of the maximum value for belief distribution with the max Deng entropy

Power law and dimension of the maximum value for belief distribution with the max Deng entropy Power law and dmenson of the maxmum value for belef dstrbuton wth the max Deng entropy Bngy Kang a, a College of Informaton Engneerng, Northwest A&F Unversty, Yanglng, Shaanx, 712100, Chna. Abstract Deng

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

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 information

Uncertainty as the Overlap of Alternate Conditional Distributions

Uncertainty as the Overlap of Alternate Conditional Distributions Uncertanty as the Overlap of Alternate Condtonal Dstrbutons Olena Babak and Clayton V. Deutsch Centre for Computatonal Geostatstcs Department of Cvl & Envronmental Engneerng Unversty of Alberta An mportant

More information

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method

Comparison 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 information

Chapter 13: Multiple Regression

Chapter 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 information

Department of Computer Science Artificial Intelligence Research Laboratory. Iowa State University MACHINE LEARNING

Department 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 information

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests Smulated of the Cramér-von Mses Goodness-of-Ft Tests Steele, M., Chaselng, J. and 3 Hurst, C. School of Mathematcal and Physcal Scences, James Cook Unversty, Australan School of Envronmental Studes, Grffth

More information

Cokriging Partial Grades - Application to Block Modeling of Copper Deposits

Cokriging Partial Grades - Application to Block Modeling of Copper Deposits Cokrgng Partal Grades - Applcaton to Block Modelng of Copper Deposts Serge Séguret 1, Julo Benscell 2 and Pablo Carrasco 2 Abstract Ths work concerns mneral deposts made of geologcal bodes such as breccas

More information

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers

Psychology 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 information

Bayesian predictive Configural Frequency Analysis

Bayesian predictive Configural Frequency Analysis Psychologcal Test and Assessment Modelng, Volume 54, 2012 (3), 285-292 Bayesan predctve Confgural Frequency Analyss Eduardo Gutérrez-Peña 1 Abstract Confgural Frequency Analyss s a method for cell-wse

More information

Comparison of Regression Lines

Comparison 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 information

Chapter 2 A Class of Robust Solution for Linear Bilevel Programming

Chapter 2 A Class of Robust Solution for Linear Bilevel Programming Chapter 2 A Class of Robust Soluton for Lnear Blevel Programmng Bo Lu, Bo L and Yan L Abstract Under the way of the centralzed decson-makng, the lnear b-level programmng (BLP) whose coeffcents are supposed

More information

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification

2E 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 information

Supplementary Notes for Chapter 9 Mixture Thermodynamics

Supplementary Notes for Chapter 9 Mixture Thermodynamics Supplementary Notes for Chapter 9 Mxture Thermodynamcs Key ponts Nne major topcs of Chapter 9 are revewed below: 1. Notaton and operatonal equatons for mxtures 2. PVTN EOSs for mxtures 3. General effects

More information

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

COMPARISON 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 information

CHAPTER-5 INFORMATION MEASURE OF FUZZY MATRIX AND FUZZY BINARY RELATION

CHAPTER-5 INFORMATION MEASURE OF FUZZY MATRIX AND FUZZY BINARY RELATION CAPTER- INFORMATION MEASURE OF FUZZY MATRI AN FUZZY BINARY RELATION Introducton The basc concept of the fuzz matr theor s ver smple and can be appled to socal and natural stuatons A branch of fuzz matr

More information

Report on Image warping

Report 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 information

A Robust Method for Calculating the Correlation Coefficient

A 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 information

Complement of Type-2 Fuzzy Shortest Path Using Possibility Measure

Complement of Type-2 Fuzzy Shortest Path Using Possibility Measure Intern. J. Fuzzy Mathematcal rchve Vol. 5, No., 04, 9-7 ISSN: 30 34 (P, 30 350 (onlne Publshed on 5 November 04 www.researchmathsc.org Internatonal Journal of Complement of Type- Fuzzy Shortest Path Usng

More information

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U)

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U) Econ 413 Exam 13 H ANSWERS Settet er nndelt 9 deloppgaver, A,B,C, som alle anbefales å telle lkt for å gøre det ltt lettere å stå. Svar er gtt . Unfortunately, there s a prntng error n the hnt of

More information

The Order Relation and Trace Inequalities for. Hermitian Operators

The Order Relation and Trace Inequalities for. Hermitian Operators Internatonal Mathematcal Forum, Vol 3, 08, no, 507-57 HIKARI Ltd, wwwm-hkarcom https://doorg/0988/mf088055 The Order Relaton and Trace Inequaltes for Hermtan Operators Y Huang School of Informaton Scence

More information

x i1 =1 for all i (the constant ).

x i1 =1 for all i (the constant ). Chapter 5 The Multple Regresson Model Consder an economc model where the dependent varable s a functon of K explanatory varables. The economc model has the form: y = f ( x,x,..., ) xk Approxmate ths by

More information

MIMA Group. Chapter 2 Bayesian Decision Theory. School of Computer Science and Technology, Shandong University. Xin-Shun SDU

MIMA Group. Chapter 2 Bayesian Decision Theory. School of Computer Science and Technology, Shandong University. Xin-Shun SDU Group M D L M Chapter Bayesan Decson heory Xn-Shun Xu @ SDU School of Computer Scence and echnology, Shandong Unversty Bayesan Decson heory Bayesan decson theory s a statstcal approach to data mnng/pattern

More information

Kernel Methods and SVMs Extension

Kernel 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 information

A New Evolutionary Computation Based Approach for Learning Bayesian Network

A 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 information

ENG 8801/ Special Topics in Computer Engineering: Pattern Recognition. Memorial University of Newfoundland Pattern Recognition

ENG 8801/ Special Topics in Computer Engineering: Pattern Recognition. Memorial University of Newfoundland Pattern Recognition EG 880/988 - Specal opcs n Computer Engneerng: Pattern Recognton Memoral Unversty of ewfoundland Pattern Recognton Lecture 7 May 3, 006 http://wwwengrmunca/~charlesr Offce Hours: uesdays hursdays 8:30-9:30

More information

MAXIMUM A POSTERIORI TRANSDUCTION

MAXIMUM 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 information

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography

CSci 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 information

The Gaussian classifier. Nuno Vasconcelos ECE Department, UCSD

The Gaussian classifier. Nuno Vasconcelos ECE Department, UCSD he Gaussan classfer Nuno Vasconcelos ECE Department, UCSD Bayesan decson theory recall that we have state of the world X observatons g decson functon L[g,y] loss of predctng y wth g Bayes decson rule s

More information

2016 Wiley. Study Session 2: Ethical and Professional Standards Application

2016 Wiley. Study Session 2: Ethical and Professional Standards Application 6 Wley Study Sesson : Ethcal and Professonal Standards Applcaton LESSON : CORRECTION ANALYSIS Readng 9: Correlaton and Regresson LOS 9a: Calculate and nterpret a sample covarance and a sample correlaton

More information

Global Sensitivity. Tuesday 20 th February, 2018

Global Sensitivity. Tuesday 20 th February, 2018 Global Senstvty Tuesday 2 th February, 28 ) Local Senstvty Most senstvty analyses [] are based on local estmates of senstvty, typcally by expandng the response n a Taylor seres about some specfc values

More information

Generative classification models

Generative classification models CS 675 Intro to Machne Learnng Lecture Generatve classfcaton models Mlos Hauskrecht mlos@cs.ptt.edu 539 Sennott Square Data: D { d, d,.., dn} d, Classfcaton represents a dscrete class value Goal: learn

More information

HLW. Vol.9 No.2 [2,3] aleatory uncertainty 10. ignorance epistemic uncertainty. variability. variability ignorance. Keywords:

HLW. Vol.9 No.2 [2,3] aleatory uncertainty 10. ignorance epistemic uncertainty. variability. variability ignorance. Keywords: Vol.9 No.2 HLW 1 varablty2 gnorance varablty gnorance 1 2 2 Keywords: Safety assessment for geologcal dsposal of hgh level radoactve waste nevtably nvolves factors that cannot be specfed n a determnstc

More information

Chapter 8 Indicator Variables

Chapter 8 Indicator Variables Chapter 8 Indcator Varables In general, e explanatory varables n any regresson analyss are assumed to be quanttatve n nature. For example, e varables lke temperature, dstance, age etc. are quanttatve n

More information

An Application of Fuzzy Hypotheses Testing in Radar Detection

An Application of Fuzzy Hypotheses Testing in Radar Detection Proceedngs of the th WSES Internatonal Conference on FUZZY SYSEMS n pplcaton of Fuy Hypotheses estng n Radar Detecton.K.ELSHERIF, F.M.BBDY, G.M.BDELHMID Department of Mathematcs Mltary echncal Collage

More information

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results.

For 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 information

5. POLARIMETRIC SAR DATA CLASSIFICATION

5. POLARIMETRIC SAR DATA CLASSIFICATION Polarmetrc SAR data Classfcaton 5. POLARIMETRIC SAR DATA CLASSIFICATION 5.1 Classfcaton of polarmetrc scatterng mechansms - Drect nterpretaton of decomposton results - Cameron classfcaton - Lee classfcaton

More information

UNIVERSITY OF TORONTO Faculty of Arts and Science. December 2005 Examinations STA437H1F/STA1005HF. Duration - 3 hours

UNIVERSITY OF TORONTO Faculty of Arts and Science. December 2005 Examinations STA437H1F/STA1005HF. Duration - 3 hours UNIVERSITY OF TORONTO Faculty of Arts and Scence December 005 Examnatons STA47HF/STA005HF Duraton - hours AIDS ALLOWED: (to be suppled by the student) Non-programmable calculator One handwrtten 8.5'' x

More information

Statistical Foundations of Pattern Recognition

Statistical Foundations of Pattern Recognition Statstcal Foundatons of Pattern Recognton Learnng Objectves Bayes Theorem Decson-mang Confdence factors Dscrmnants The connecton to neural nets Statstcal Foundatons of Pattern Recognton NDE measurement

More information

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Chapter 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 information

Orientation Model of Elite Education and Mass Education

Orientation Model of Elite Education and Mass Education Proceedngs of the 8th Internatonal Conference on Innovaton & Management 723 Orentaton Model of Elte Educaton and Mass Educaton Ye Peng Huanggang Normal Unversty, Huanggang, P.R.Chna, 438 (E-mal: yepeng@hgnc.edu.cn)

More information

Multiple Sound Source Location in 3D Space with a Synchronized Neural System

Multiple Sound Source Location in 3D Space with a Synchronized Neural System Multple Sound Source Locaton n D Space wth a Synchronzed Neural System Yum Takzawa and Atsush Fukasawa Insttute of Statstcal Mathematcs Research Organzaton of Informaton and Systems 0- Mdor-cho, Tachkawa,

More information

Computation of Higher Order Moments from Two Multinomial Overdispersion Likelihood Models

Computation of Higher Order Moments from Two Multinomial Overdispersion Likelihood Models Computaton of Hgher Order Moments from Two Multnomal Overdsperson Lkelhood Models BY J. T. NEWCOMER, N. K. NEERCHAL Department of Mathematcs and Statstcs, Unversty of Maryland, Baltmore County, Baltmore,

More information

SPANC -- SPlitpole ANalysis Code User Manual

SPANC -- SPlitpole ANalysis Code User Manual Functonal Descrpton of Code SPANC -- SPltpole ANalyss Code User Manual Author: Dale Vsser Date: 14 January 00 Spanc s a code created by Dale Vsser for easer calbratons of poston spectra from magnetc spectrometer

More information

Generalized Linear Methods

Generalized 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 information

PATTERN RECOGNITION AND IMAGE UNDERSTANDING

PATTERN RECOGNITION AND IMAGE UNDERSTANDING PATTERN RECOGNITION AND IMAGE UNDERSTANDING The ultmate objectve of many mage analyss tasks s to dscover meanng of the analysed mage, e.g. categorse the objects, provde symbolc/semantc nterpretaton of

More information

INF 5860 Machine learning for image classification. Lecture 3 : Image classification and regression part II Anne Solberg January 31, 2018

INF 5860 Machine learning for image classification. Lecture 3 : Image classification and regression part II Anne Solberg January 31, 2018 INF 5860 Machne learnng for mage classfcaton Lecture 3 : Image classfcaton and regresson part II Anne Solberg January 3, 08 Today s topcs Multclass logstc regresson and softma Regularzaton Image classfcaton

More information

Regularized Discriminant Analysis for Face Recognition

Regularized 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 information

CONTRAST ENHANCEMENT FOR MIMIMUM MEAN BRIGHTNESS ERROR FROM HISTOGRAM PARTITIONING INTRODUCTION

CONTRAST 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 information

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur Analyss of Varance and Desgn of Experment-I MODULE VII LECTURE - 3 ANALYSIS OF COVARIANCE Dr Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur Any scentfc experment s performed

More information

Clustering & Unsupervised Learning

Clustering & Unsupervised Learning Clusterng & Unsupervsed Learnng Ken Kreutz-Delgado (Nuno Vasconcelos) ECE 175A Wnter 2012 UCSD Statstcal Learnng Goal: Gven a relatonshp between a feature vector x and a vector y, and d data samples (x,y

More information

Composite Hypotheses testing

Composite 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 information

Internet Engineering. Jacek Mazurkiewicz, PhD Softcomputing. Part 3: Recurrent Artificial Neural Networks Self-Organising Artificial Neural Networks

Internet Engineering. Jacek Mazurkiewicz, PhD Softcomputing. Part 3: Recurrent Artificial Neural Networks Self-Organising Artificial Neural Networks Internet Engneerng Jacek Mazurkewcz, PhD Softcomputng Part 3: Recurrent Artfcal Neural Networks Self-Organsng Artfcal Neural Networks Recurrent Artfcal Neural Networks Feedback sgnals between neurons Dynamc

More information

Statistics II Final Exam 26/6/18

Statistics II Final Exam 26/6/18 Statstcs II Fnal Exam 26/6/18 Academc Year 2017/18 Solutons Exam duraton: 2 h 30 mn 1. (3 ponts) A town hall s conductng a study to determne the amount of leftover food produced by the restaurants n the

More information

Homework Assignment 3 Due in class, Thursday October 15

Homework Assignment 3 Due in class, Thursday October 15 Homework Assgnment 3 Due n class, Thursday October 15 SDS 383C Statstcal Modelng I 1 Rdge regresson and Lasso 1. Get the Prostrate cancer data from http://statweb.stanford.edu/~tbs/elemstatlearn/ datasets/prostate.data.

More information

Numerical Heat and Mass Transfer

Numerical 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 information

Conjugacy and the Exponential Family

Conjugacy and the Exponential Family CS281B/Stat241B: Advanced Topcs n Learnng & Decson Makng Conjugacy and the Exponental Famly Lecturer: Mchael I. Jordan Scrbes: Bran Mlch 1 Conjugacy In the prevous lecture, we saw conjugate prors for the

More information

The Quadratic Trigonometric Bézier Curve with Single Shape Parameter

The Quadratic Trigonometric Bézier Curve with Single Shape Parameter J. Basc. Appl. Sc. Res., (3541-546, 01 01, TextRoad Publcaton ISSN 090-4304 Journal of Basc and Appled Scentfc Research www.textroad.com The Quadratc Trgonometrc Bézer Curve wth Sngle Shape Parameter Uzma

More information

Logistic Regression. CAP 5610: Machine Learning Instructor: Guo-Jun QI

Logistic 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 information

ERROR RATES STABILITY OF THE HOMOSCEDASTIC DISCRIMINANT FUNCTION

ERROR RATES STABILITY OF THE HOMOSCEDASTIC DISCRIMINANT FUNCTION ISSN - 77-0593 UNAAB 00 Journal of Natural Scences, Engneerng and Technology ERROR RATES STABILITY OF THE HOMOSCEDASTIC DISCRIMINANT FUNCTION A. ADEBANJI, S. NOKOE AND O. IYANIWURA 3 *Department of Mathematcs,

More information

Solution Thermodynamics

Solution Thermodynamics Soluton hermodynamcs usng Wagner Notaton by Stanley. Howard Department of aterals and etallurgcal Engneerng South Dakota School of nes and echnology Rapd Cty, SD 57701 January 7, 001 Soluton hermodynamcs

More information

Improvement of Histogram Equalization for Minimum Mean Brightness Error

Improvement of Histogram Equalization for Minimum Mean Brightness Error Proceedngs of the 7 WSEAS Int. Conference on Crcuts, Systems, Sgnal and elecommuncatons, Gold Coast, Australa, January 7-9, 7 3 Improvement of Hstogram Equalzaton for Mnmum Mean Brghtness Error AAPOG PHAHUA*,

More information

Mixture o f of Gaussian Gaussian clustering Nov

Mixture o f of Gaussian Gaussian clustering Nov Mture of Gaussan clusterng Nov 11 2009 Soft vs hard lusterng Kmeans performs Hard clusterng: Data pont s determnstcally assgned to one and only one cluster But n realty clusters may overlap Soft-clusterng:

More information

Solving Nonlinear Differential Equations by a Neural Network Method

Solving 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 information

Turbulence classification of load data by the frequency and severity of wind gusts. Oscar Moñux, DEWI GmbH Kevin Bleibler, DEWI GmbH

Turbulence classification of load data by the frequency and severity of wind gusts. Oscar Moñux, DEWI GmbH Kevin Bleibler, DEWI GmbH Turbulence classfcaton of load data by the frequency and severty of wnd gusts Introducton Oscar Moñux, DEWI GmbH Kevn Blebler, DEWI GmbH Durng the wnd turbne developng process, one of the most mportant

More information

Checking Pairwise Relationships. Lecture 19 Biostatistics 666

Checking Pairwise Relationships. Lecture 19 Biostatistics 666 Checkng Parwse Relatonshps Lecture 19 Bostatstcs 666 Last Lecture: Markov Model for Multpont Analyss X X X 1 3 X M P X 1 I P X I P X 3 I P X M I 1 3 M I 1 I I 3 I M P I I P I 3 I P... 1 IBD states along

More information

The Expectation-Maximization Algorithm

The Expectation-Maximization Algorithm The Expectaton-Maxmaton Algorthm Charles Elan elan@cs.ucsd.edu November 16, 2007 Ths chapter explans the EM algorthm at multple levels of generalty. Secton 1 gves the standard hgh-level verson of the algorthm.

More information

Tensor Smooth Length for SPH Modelling of High Speed Impact

Tensor Smooth Length for SPH Modelling of High Speed Impact Tensor Smooth Length for SPH Modellng of Hgh Speed Impact Roman Cherepanov and Alexander Gerasmov Insttute of Appled mathematcs and mechancs, Tomsk State Unversty 634050, Lenna av. 36, Tomsk, Russa RCherepanov82@gmal.com,Ger@npmm.tsu.ru

More information

Comparative Studies of Law of Conservation of Energy. and Law Clusters of Conservation of Generalized Energy

Comparative Studies of Law of Conservation of Energy. and Law Clusters of Conservation of Generalized Energy Comparatve Studes of Law of Conservaton of Energy and Law Clusters of Conservaton of Generalzed Energy No.3 of Comparatve Physcs Seres Papers Fu Yuhua (CNOOC Research Insttute, E-mal:fuyh1945@sna.com)

More information

Relevance Vector Machines Explained

Relevance Vector Machines Explained October 19, 2010 Relevance Vector Machnes Explaned Trstan Fletcher www.cs.ucl.ac.uk/staff/t.fletcher/ Introducton Ths document has been wrtten n an attempt to make Tppng s [1] Relevance Vector Machnes

More information

A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS

A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS HCMC Unversty of Pedagogy Thong Nguyen Huu et al. A PROBABILITY-DRIVEN SEARCH ALGORITHM FOR SOLVING MULTI-OBJECTIVE OPTIMIZATION PROBLEMS Thong Nguyen Huu and Hao Tran Van Department of mathematcs-nformaton,

More information

Inductance Calculation for Conductors of Arbitrary Shape

Inductance 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 information

Quantum and Classical Information Theory with Disentropy

Quantum and Classical Information Theory with Disentropy Quantum and Classcal Informaton Theory wth Dsentropy R V Ramos rubensramos@ufcbr Lab of Quantum Informaton Technology, Department of Telenformatc Engneerng Federal Unversty of Ceara - DETI/UFC, CP 6007

More information

Errors in Nobel Prize for Physics (7) Improper Schrodinger Equation and Dirac Equation

Errors in Nobel Prize for Physics (7) Improper Schrodinger Equation and Dirac Equation Errors n Nobel Prze for Physcs (7) Improper Schrodnger Equaton and Drac Equaton u Yuhua (CNOOC Research Insttute, E-mal:fuyh945@sna.com) Abstract: One of the reasons for 933 Nobel Prze for physcs s for

More information

Measurement Indices of Positional Uncertainty for Plane Line Segments Based on the ε

Measurement Indices of Positional Uncertainty for Plane Line Segments Based on the ε Proceedngs of the 8th Internatonal Smposum on Spatal ccurac ssessment n Natural Resources and Envronmental Scences Shangha, P R Chna, June 5-7, 008, pp 9-5 Measurement Indces of Postonal Uncertant for

More information

x = , so that calculated

x = , 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 information

Semi-supervised Classification with Active Query Selection

Semi-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 information

FUZZY GOAL PROGRAMMING VS ORDINARY FUZZY PROGRAMMING APPROACH FOR MULTI OBJECTIVE PROGRAMMING PROBLEM

FUZZY 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 information

Color Rendering Uncertainty

Color Rendering Uncertainty Australan Journal of Basc and Appled Scences 4(10): 4601-4608 010 ISSN 1991-8178 Color Renderng Uncertanty 1 A.el Bally M.M. El-Ganany 3 A. Al-amel 1 Physcs Department Photometry department- NIS Abstract:

More information

XII.3 The EM (Expectation-Maximization) Algorithm

XII.3 The EM (Expectation-Maximization) Algorithm XII.3 The EM (Expectaton-Maxzaton) Algorth Toshnor Munaata 3/7/06 The EM algorth s a technque to deal wth varous types of ncoplete data or hdden varables. It can be appled to a wde range of learnng probles

More information

On mutual information estimation for mixed-pair random variables

On mutual information estimation for mixed-pair random variables On mutual nformaton estmaton for mxed-par random varables November 3, 218 Aleksandr Beknazaryan, Xn Dang and Haln Sang 1 Department of Mathematcs, The Unversty of Msssspp, Unversty, MS 38677, USA. E-mal:

More information

Markov Chain Monte Carlo (MCMC), Gibbs Sampling, Metropolis Algorithms, and Simulated Annealing Bioinformatics Course Supplement

Markov Chain Monte Carlo (MCMC), Gibbs Sampling, Metropolis Algorithms, and Simulated Annealing Bioinformatics Course Supplement Markov Chan Monte Carlo MCMC, Gbbs Samplng, Metropols Algorthms, and Smulated Annealng 2001 Bonformatcs Course Supplement SNU Bontellgence Lab http://bsnuackr/ Outlne! Markov Chan Monte Carlo MCMC! Metropols-Hastngs

More information

CHALMERS, GÖTEBORGS UNIVERSITET. SOLUTIONS to RE-EXAM for ARTIFICIAL NEURAL NETWORKS. COURSE CODES: FFR 135, FIM 720 GU, PhD

CHALMERS, GÖTEBORGS UNIVERSITET. SOLUTIONS to RE-EXAM for ARTIFICIAL NEURAL NETWORKS. COURSE CODES: FFR 135, FIM 720 GU, PhD CHALMERS, GÖTEBORGS UNIVERSITET SOLUTIONS to RE-EXAM for ARTIFICIAL NEURAL NETWORKS COURSE CODES: FFR 35, FIM 72 GU, PhD Tme: Place: Teachers: Allowed materal: Not allowed: January 2, 28, at 8 3 2 3 SB

More information

Lecture Notes on Linear Regression

Lecture 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 information

DUE: WEDS FEB 21ST 2018

DUE: WEDS FEB 21ST 2018 HOMEWORK # 1: FINITE DIFFERENCES IN ONE DIMENSION DUE: WEDS FEB 21ST 2018 1. Theory Beam bendng s a classcal engneerng analyss. The tradtonal soluton technque makes smplfyng assumptons such as a constant

More information

Multilayer Perceptron (MLP)

Multilayer 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 information

Hidden Markov Models & The Multivariate Gaussian (10/26/04)

Hidden Markov Models & The Multivariate Gaussian (10/26/04) CS281A/Stat241A: Statstcal Learnng Theory Hdden Markov Models & The Multvarate Gaussan (10/26/04) Lecturer: Mchael I. Jordan Scrbes: Jonathan W. Hu 1 Hdden Markov Models As a bref revew, hdden Markov models

More information

Excess Error, Approximation Error, and Estimation Error

Excess Error, Approximation Error, and Estimation Error E0 370 Statstcal Learnng Theory Lecture 10 Sep 15, 011 Excess Error, Approxaton Error, and Estaton Error Lecturer: Shvan Agarwal Scrbe: Shvan Agarwal 1 Introducton So far, we have consdered the fnte saple

More information

Laboratory 1c: Method of Least Squares

Laboratory 1c: Method of Least Squares Lab 1c, Least Squares Laboratory 1c: Method of Least Squares Introducton Consder the graph of expermental data n Fgure 1. In ths experment x s the ndependent varable and y the dependent varable. Clearly

More information