Neurocomputing. Kernel sparse representation based classification. Jun Yin a,n, Zhonghua Liu a, Zhong Jin a, Wankou Yang b. abstract.

Size: px
Start display at page:

Download "Neurocomputing. Kernel sparse representation based classification. Jun Yin a,n, Zhonghua Liu a, Zhong Jin a, Wankou Yang b. abstract."

Transcription

1 Neurocomputing 77 (2012) Contents lists ville t SciVerse ScienceDirect Neurocomputing journl homepge: Kernel sprse representtion sed clssifiction Jun Yin,n, Zhonghu Liu, Zhong Jin, Wnkou Yng School of Computer Science nd Technology, Nnjing University of Science nd Technology, Nnjing , Chin School of Automtion, Southest University, Nnjing , Chin rticle info Article history: Received 22 Decemer 2010 Received in revised form 23 August 2011 Accepted 28 August 2011 Communicted y Y. Fu Aville online 22 Septemer 2011 Keywords: Clssifiction Sprse representtion Kernel strct Sprse representtion hs ttrcted gret ttention in the pst few yers. Sprse representtion sed clssifiction (SRC) lgorithm ws developed nd successfully used for clssifiction. In this pper, kernel sprse representtion sed clssifiction (KSRC) lgorithm is proposed. Smples re mpped into high dimensionl feture spce first nd then SRC is performed in this new feture spce y utilizing kernel trick. Since smples in the high dimensionl feture spce re unknown, we cnnot perform KSRC directly. In order to overcome this difficulty, we give the method to solve the prolem of sprse representtion in the high dimensionl feture spce. If n pproprite kernel is selected, in the high dimensionl feture spce, test smple is proly represented s the liner comintion of trining smples of the sme clss more ccurtely. Therefore, KSRC hs more powerful clssifiction ility thn SRC. Experiments of fce recognition, plmprint recognition nd finger-knuckle-print recognition demonstrte the effectiveness of KSRC. & 2011 Elsevier B.V. All rights reserved. 1. Introduction In pttern recognition, clssifiction is n indispensle step nd clssifier design is one of the most populr technologies. Severl clssifiction pproches hve een proposed over the pst severl decdes [1,2]. Among them, the nerest-neighor (NN) clssifier nd the nerest-men (NM) clssifier re most widely used ecuse of their simpleness nd vilility. The NN clssifier ssigns test smple to the ctegory of its nerest neighor from the leled trining set. Insted of serching the nerest trining smple, the NM clssifier ssigns test smple to the ctegory of its nerest clss men. Over the lst more thn 10 yers, the kernel sed lgorithms [3] such s kernel principl component nlysis (KPCA) [4] nd kernel fisher discriminnt nlysis (KFD) [5,6] hve roused considerle interest in pttern recognition nd mchine lerning. These kernel sed lgorithms improve the computtionl power of the liner lgorithms. They mp the dt into high dimensionl feture spce y nonliner mpping nd perform liner lgorithms in the high dimensionl feture spce using the inner products. In the high dimensionl feture spce, the inner products cn e computed y kernel function. For clssifiction, utilizing kernel pproch, Yu et l. present the kernel nerest neighor (KERNEL-NN) clssifier [7]. KERNEL-NN pplies the nerest neighor clssifiction method in the high dimensionl n Corresponding uthor. E-mil ddress: yinjun8429@163.com (J. Yin). feture spce. Kernel pproch could chnge the distriution of smples y the nonliner mpping. Some linerly inseprle smples in the originl feture spce cn ecome linerly seprle in the high dimensionl feture spce. If n pproprite kernel is chosen to reshpe the distriution of smples, the KERNEL-NN clssifier could perform etter thn the NN clssifier. Recently, sprse representtion ecomes hot topic of pttern recognition nd computer vision. It is pplied to imge superresolution [8], motion segmenttion [9] nd supervised denoising [10]. Wright et l. pply sprse representtion to clssifiction nd exploit the sprse representtion sed clssifiction (SRC) lgorithm [11]. For SRC, test smple is represented s sprse comintion of trining smples, nd its sprse representtion coefficient is otined y solving the prolem of sprse representtion. The test smple is ssigned to the clss tht minimizes the residul etween itself nd the reconstruction constructed y trining smples of this clss. SRC shows its effectiveness in fce recognition experiments. As well known, fter mpping smples into high dimensionl feture spce y nonliner mpping, kernel pproch cn chnge the distriution of smples. If n pproprite kernel function is utilized, for test smple, more neighors proly hve the sme clss lel s itself in the high dimensionl feture spce. Here, in the high dimensionl feture spce, the test smple cn e represented more ccurtely s the comintion of trining smples from the sme clss. Then the nonzero entries of sprse representtion coefficient vector of the test smple will e more ssocited with trining smples from the sme clss s itself. Nmely, sprse representtion coefficient in the high dimensionl /$ - see front mtter & 2011 Elsevier B.V. All rights reserved. doi: /j.neucom

2 J. Yin et l. / Neurocomputing 77 (2012) feture spce cn denote the ctegory of the test smple more ccurtely nd it hs more powerful discriminting ility. To use sprse representtion coefficient in the high dimensionl feture spce, we propose the kernel sprse representtion sed clssifiction (KSRC) lgorithm in this pper. For KSRC, smples re mpped into high dimensionl feture spce first nd then SRC is performed in this new feture spce. We prove tht SRC in the high dimensionl feture spce cn e formulted in terms of the inner products, while the inner products could e computed y kernel function. Comprehensive comprisons etween KSRC nd NM, NN, KERNEL-NN nd SRC revel the superior chrcteristics of KSRC. The rest of the pper is orgnized s follows: Section 2 descries SRC lgorithm nd proposes KSRC lgorithm. Section 3 descries experiments on severl populr dtses. Finlly, the conclusions re summrized in Section The proposed lgorithm 2.1. Sprse representtion sed clssifiction (SRC) [11] Suppose tht we hve n trining smples for c clsses nd sufficient trining smples elong to the kth clss, A k ¼½x k,1, x k,2,...,x k,nk ŠAR mn k, where m is the dimension of smples nd n k is the numer of trining smples of the kth clss. Any test smple yar m from the kth clss cn e pproximtely represented s the liner comintion of trining smples of this clss: y ¼ k,1 x k,1 þ k,2 x k,2 þþ k,nk x k,nk ð1þ Since the lel of y is unknown initilly, we represent y s the liner comintion of ll the trining smples: y ¼ A 0 ð2þ where A ¼½A 1,A 2,...,A c Š¼½x 1,1,x 1,2,...,x c,nc ŠAR mn is mtrix composed of ll the n trining smples of c clsses nd 0 ¼ ½0,...,0, k,1, k,2,..., k,nk,0,...,0š T AR n is the coefficient vector whose nonzero entries re only ssocited with the kth clss. When c is lrge, 0 will e sprse. If mon, Eq. (2) is underdetermined. The prolem of sprse representtion is to serch vector such tht Eq. (2) is stisfied nd :: 0 is minimized, where :: 0 is the l 0 -norm of. This cn e descried s ^ 0 ¼ rg min :: 0 suject to y ¼ A ð3þ However, finding the sprse solution of Eq.(3) is NP-hrd [12]: nmely, there is no known procedure for otining the sprsest solution, which is significntly more efficient thn exhusting ll susets of the entries for. The theory of sprse representtion nd compressive sensing [13 15] revels tht we cn solve the following convex relxed optimiztion to otin pproximte solution: ^ 1 ¼ rg min :: 1 suject to y ¼ A ð4þ where :: 1 is the l 1 -norm of. Thisprolemcnesolvedy stndrd liner progrmming methods [16]. Furthermore, the oservtions re often inccurte, then we should relx the constrint in Eq. (4) nd get the following optimiztion prolem: ^ 1 ¼ rg min :: 1 sujectto :A y: 2 re ð5þ where e is the tolernce of the reconstruction error. This convex optimiztion prolem cn e solved vi second-order cone progrmming [16]. The optimiztion prolem (5) is minly used to del with smll noise. In prctice, the oservtions possily contin ig noise. For exmple, the imges re corrupted or occluded. Here, the errors cnnot e ignored or solved y the optimiztion prolem (5). The constrint should e modified s y ¼ Aþe ¼½AIŠ ð6þ e where ear m is vector of errors, IAR mm is the identity mtrix. Now, we get the following optimiztion prolem: ^g 1 ¼ rg min :g: 1 suject to y ¼ Pg ð7þ g where P ¼½A IŠAR mðn þ mþ,g ¼ " AR n þ m nd ^g ^ # 1 e 1 ¼ AR n þ m ^e 1 Let ^ 1 denote the solution of sprse representtion prolem (7) otined y l 1 -minimiztion. Idelly, the nonzero entries in ^ 1 will e ssocited with the columns of A from single oject clss, nd we cn esily ssign the test smple y to tht clss. However, noise nd modeling error my cuse smll nonzero entries ssocited with multiple clsses. Simple heuristics such s ssigning y to the clss with the lrgest entry re not dependle. Insted, we define new vector ^ k 1ðk ¼ 1,2,...,cÞ whose only nonzero entries re the entries in ^ 1 tht re ssocited with clss k. The reconstruction with the trining smples of the kth clss is ^y k ¼ A^ k 1ðk ¼ 1,2,...,cÞ. Then y cn e ssigned to the clss tht minimizes the residul etween y nd ^y k : min r k ðyþ¼:y A^ k 1 : 2 k The SRC lgorithm is summrized s follows: Algorithm 1. Sprse representtion sed clssifiction (SRC) 1. Input: the mtrix of trining smples AAR mn, test smple yar m. 2. Normlize the columns of A to hve unit l 2 -norm. 3. Solve the l 1 -minimiztion prolem defined in Eq. (4) or (5) or (7). 4. Compute the residuls r k ðyþ ðk ¼ 1,2,...,cÞ defined in Eq. (8). 5. Output : identityðyþ¼rg minðr k ðyþþ k 2.2. Kernel sprse representtion sed clssifiction (KSRC) As we know, kernel pproch cn chnge the distriution of smples y mpping smples into high dimensionl feture spce [7]. This chnge possily hs two effects if n pproprite kernel function is selected. On the one hnd, some liner inseprle smples in the originl feture spce ecome liner seprle in the high dimensionl feture spce. This leds to superiority of the KERNEL-NN clssifier over the NN clssifier. On the other hnd, test smple cn e represented s the liner comintion of trining smples from the sme clss s itself more ccurtely in the high dimensionl feture spce thn originl. Then the nonzero entries of sprse representtion coefficient vector of the test smple re more ssocited with trining smples of the sme clss. This results in etter clssifiction ility of SRC. So we perform SRC in the high dimensionl feture spce nd propose kernel sprse representtion sed clssifiction (KSRC). Becuse the explicit mpping from the originl feture spce to the high dimensionl feture spce is unknown, KSRC cnnot e performed directly. However, we successfully solve this prolem y Theorem 1. Suppose tht smples re mpped from originl feture spce R m into high dimensionl feture spce H y nonliner ð8þ

3 122 J. Yin et l. / Neurocomputing 77 (2012) mpping f: R m -H, x-fðxþ ð9þ Let B ¼½fðx 1,1 Þ,fðx 1,2 Þ,...,fðx c,nc ÞŠ represent the mtrix composed of ll the trining smples fter the nonliner mpping f. The prolem of sprse representtion in H cn e descried s [11] ^ 0 ¼ rg min :: 0 suject to fðyþ¼b ð10þ where f(y) is ny test smple in the high dimensionl feture spce, which corresponds to y in the originl feture spce. Similrly, the pproximte solution of Eq. (10) cn e otined through the following convex relxed optimiztion [13 15]: ^ 1 ¼ rg min :: 1 suject to fðyþ¼b ð11þ When the oservtions re not ccurte, the constrint in Eq. (11) should e relxed nd the following optimiztion prolem is otined: ^ 1 ¼ rg min :: 1 suject to :B fðyþ: 2 re ð12þ If we set e¼0, Eq. (12) is equivlent to Eq. (11). So Eq. (11) cn e seen s specil cse of Eq. (12) nd we cn only consider the optimiztion prolem (12). Since B nd f(y) re unknown, Eq. (12) cnnot e solved directly. But ccording to Theorem 1, Eq. (12) cn e trnsformed to ^ 1 ¼ rg min :: 1 suject to :B T B B T fðyþ: 2 rd ð13þ Theorem 1. For ny ez0, there must exist dz0 such tht we hve :B fðyþ: 2 re, s long s :B T B B T fðyþ: 2 rd is stisfied. The inner product of smples in the high dimensionl feture spce cn e computed y kernel function. Nmely, for ny smples x nd y, we hve fðxþ T fðyþ¼kðx,yþ, where kðx,yþ is kernel function. Then B T B ¼½fðx 1,1 Þ,fðx 1,2 Þ,...,fðx c,nc ÞŠ T ½fðx 1,1 Þ,fðx 1,2 Þ,...,fðx c,nc ÞŠ 2 3 kðx 1,1,x 1,1 Þ kðx 1,1,x 1,2 Þ kðx 1,1,x c,nc Þ kðx 1,2,x 1,1 Þ kðx 1,2,x 1,2 Þ kðx 1,2,x c,nc Þ ¼ 6 ^ ^ & ^ kðx c,nc,x 1,1 Þ kðx c,nc,x 1,2 Þ kðx c,nc,x c,nc Þ nd 2 3 kðx 1,1,yÞ B T kðx fðyþ¼½fðx 1,1 Þ,fðx 1,2 Þ,...,fðx c,nc ÞŠ T 1,2,yÞ fðyþ¼ 6 4 ^ 7 5 kðx c,nc,yþ ð14þ ð15þ When the kernel function kðx,yþ is given, B T B nd B T fðyþ re otined. Now we could solve the convex optimiztion prolem (13) vi second-order cone progrmming [16]. If the oservtions contin ig noise, s SRC, the constrint in Eq. (13) should e modified s B T fðyþ¼b T BþE ¼½B T B IŠ ~ ð16þ E where EAR n is vector of errors, I ~ AR nn is the identity mtrix. Utilizing constrint (16), the following optimiztion prolem is otined: ^Z 1 ¼ rg min :Z: 1 suject to B T fðyþ¼qz ð17þ Z where Q ¼½B T B IŠAR ~ n2n,z ¼ " # ^ AR 2n nd ^Z 1 E 1 ¼ ^E 1 AR 2n Let ^ 1 denote the solution of optimiztion prolem (17). Similr to SRC, we define new vector ^ k 1ðk ¼ 1,2,...,cÞ y setting only those entries in ^ 1 ssocited with clss k nonzero nd ssigning zero to other entries. Then y cn e ssigned to the clss tht minimizes the residul etween B T fðyþ nd B T B ^ k 1 : min R k ðyþ¼:b T fðyþ B T B ^ k 1 : 2 k ð18þ Algorithm 2. Kernel sprse representtion sed clssifiction (KSRC) 1. Input: the mtrix of trining smples AAR mn, test smple yar m nd kernel function. 2. Normlize the columns of A to hve unit l 2 -norm. 3. Clculte B T B nd B T fðyþ y Eqs. (14) nd (15). 4. Solve the l 1 -minimiztion prolem defined in Eqs. (13) or (17). 5. Compute the residuls R k ðyþ ðk ¼ 1,2,...,cÞ defined in Eq. (18). 6. Output : identityðyþ¼rg minðr k ðyþþ k For smples contining smll noise, the computtionl cost of SRC nd KSRC is minly cused y solving the convex optimiztion prolem (5) nd (13), respectively. According to AAR mn nd B T BAR nn, the computtionl complexity of solving Eqs. (5) nd (13) re oth Oðn 3 Þ. Here, SRC nd KSRC hve the sme computtionl cost. For smples contining ig noise, the computtionl cost of SRC nd KSRC is minly cused y solving the convex optimiztion prolems (7) nd (17) seprtely. We know tht the size of P is m ðnþmþ nd the size of Q is n 2n. Then the computtionl complexity of solving Eq. (7) is OððnþmÞ 3 Þ nd the computtionl complexity of solving Eq. (17) is Oðð2nÞ 3 Þ [23]. At this time, if the numer of trining smple size n is smller thn the dimension m, the computtionl cost of KSRC is shorter thn SRC. Otherwise, the computtionl cost of KSRC is longer thn SRC. 3. Experiments In this section, the effectiveness of KSRC lgorithm is evluted y experiments. We do experiments on FERET, ORL, Yle nd AR fce dtses nd the PolyU plmprint nd finger-knuckle-print (FKP) dtses. Principl component nlysis (PCA) [17] nd rndom projection (RP) [18] re used for feture extrction. We compre the clssifiction ility of KSRC lgorithm with NM, NN, KERNEL-NN nd SRC lgorithms fter feture extrction. Two populr kernels re involved in our experiments. One is polynomil kernel kðx,yþ¼ð1þx T yþ d nd the other is Gussin kernel kðx,yþ¼ expð :x y: 2 =tþ. For KSRC, we use these two kernels, respectively. Since KERNEL-NN [7] using Gussin kernel is equivlent to NN, only polynomil kernel is used for KERNEL-NN. The optiml kernel prmeters re selected Dt corpor FERET fce dtse The FERET fce dtse [19] ws sponsored y the US Deprtment of Defense through the DARPA Progrm. It hs ecome stndrd dtse for testing nd evluting fce recognition lgorithms. We perform lgorithms on suset of

4 J. Yin et l. / Neurocomputing 77 (2012) the FERET fce dtse. The suset is composed of 1400 imges of 200 individuls, nd ech individul hs seven imges. It involves vritions in fce expression, pose nd illumintion. In the experiment, the fcil portion of the originl imge ws cropped sed on the loction of eyes nd mouth. Then we resized the cropped imges to pixels nd preprocess them y histogrm equliztion. Seven smple imges of one person re shown in Fig ORL fce dtse ORL fce dtse contins 400 fce imges of 40 individuls. The imge size is with 256 gry levels per pixel. The fce imges re centrlized. There re vritions in pose, illumintion nd fcil expression. Fig. 2 shows smple imges of one person Yle fce dtse The Yle fce dtse ws constructed t the Yle Center for Computtionl Vision nd Control. It contins 165 gry-scle imges of 15 individuls. The imges demonstrte vritions in lighting, fcil expression nd with/without glsses. In our experiment, every imge ws mnully cropped nd resized to pixels. Fig. 3 shows 11 imges of one people AR fce dtse The AR fce dtse [20] contins over 4000 color fce imges of 126 people, including 26 frontl views of fces with different fcil expressions, lighting conditions, nd occlusions for ech people. The pictures of 120 individuls were collected in two sessions (14 dys prt) nd ech session contins 13 color imges. Fourteen fce imges (ech session contining 7) of these 120 individuls re selected in our experiment. The imges re converted to gryscle. The fce portion of ech imge is mnully cropped nd normlized to pixels. Fig. 4 shows smple imges of one person. These imges vry s follows: neutrl expression, smiling, ngry, screming, left light on, right light on, ll sides light on PolyU FKP dtse FKP imges on the PolyU FKP dtse were collected from 165 volunteers. These imges re collected in two seprte sessions. In ech session, the suject ws sked to provide six imges for ech of the left index finger, the left middle finger, the right index finger nd the right middle finger. The imges were processed y ROI extrction lgorithm descried in [21]. In the experiment, we select 1200 FKP imges of the right index finger of 100 sujects. These selected imges were resized to pixels nd preprocessed y histogrm equliztion. Fig. 5 shows 12 smple imges of one right index finger PolyU plmprint dtse The PolyU plmprint dtse contins 600 imges of 100 different plms with six smples for ech plm. Six smples from ech of these plms were collected in two sessions, where the first three were cptured in the first session nd the other three in Fig. 1. Smple imges of one person on FERET fce dtse. Fig. 2. Smple imges of one person on ORL fce dtse.

5 124 J. Yin et l. / Neurocomputing 77 (2012) Fig. 3. Smple imges of one person on Yle fce dtse. Fig. 4. Smple imges of one person on AR fce dtse. the second session. The centrl prt of ech originl imge ws utomticlly cropped using the lgorithm mentioned in [22]. The cropped imges were resized to pixels nd preprocessed y histogrm equliztion. Fig. 6 shows six smple imges of one plm Experimentl results On FERET fce dtse, first we try to find the optiml kernel prmeters for KSRC using glol-to-locl serch strtegy [3]. Three imges per person re rndomly selected for trining nd the remining four imges re used for vlidtion. After feture extrction y PCA, the dimension of the smples is fixed t 150. Through glolly serching over wide rnge of the prmeter spce, we find cndidte intervl where the optiml prmeters my exist. Here, for the prmeter d of polynomil kernel, the cndidte intervl is from 1 to 10, for the prmeter t of Gussin kernel, the cndidte intervl is lso from 1 to 10. Now, we try to find the optiml kernel prmeters within these intervls. Fig. 7 shows the recognition rtes of KSRC with polynomil kernel versus the vrition of the prmeter d. Fig. 7 shows the recognition rtes of KSRC with Gussin kernel versus the vrition of the prmeter t. From Fig. 7 nd, we cn see tht the optiml prmeter d is 2 nd the optiml prmeter t is lso 2. After determining the optiml kernel prmeters, we compre KSRC with NM, NN, KERNEL-NN nd SRC. The first three imges per person re used for trining nd the rest four imges re used for testing. Tle 1 shows the mximl recognition rtes of five methods nd the corresponding dimensions nd prmeters. From Tle 1, it cn e seen tht KSRC outperforms other four methods, whether polynomil kernel is used or Gussin kernel is used. In the first experiment on ORL fce dtse, three imges per individul re rndomly chosen for trining nd the remining seven imges re used for testing. PCA is used for feture extrction. The experiment is repeted for 20 times. The first 10 times re used

6 J. Yin et l. / Neurocomputing 77 (2012) for tuning kernel prmeters nd the other 10 times for compring the performnce of NM, NN, KERNEL-NN, SRC nd KSRC. The optiml kernel prmeters re lso determined y glol-to-locl serch strtegy. Fig. 8 shows the verge recognition rtes versus the dimensions. Tle 2 lists the mximl verge recognition rte nd the stndrd devition of ech method cross 10 runs nd the corresponding dimension nd prmeter. From Fig. 8 nd Tle 2, we cn see four min points. First, no mtter which kernel is used, our KSRC consistently outperforms other four lgorithms irrespective of the vrition of dimensions. Second, SRC performs etter thn NM, NN nd KERNEL-NN lgorithms when the dimension is over out 20. Third, KERNEL-NN lmost outperforms NN. Lst, NM hs the worst performnce in this experiment. From the first nd the third points, we cn see tht kernel pproch indeed improve the clssifiction ility. We know SRC hs good performnce for recognition under occlusion. In the second experiment on ORL fce dtse, we test the ility of KSRC for hndling occlusion. For the lst imge of every individul, the region from 30 to 60 in width nd from 30 to 60 in length ws replced y lck lock. Fig. 9 shows the imge of one person under occlusion. We use the first nine imges per person for trining nd the lst imge under occlusion for testing. SRC nd KSRC re used for clssifiction fter PCA trnsformtion. The mximl recognition rtes nd the corresponding dimensions of two clssifiers re given in Tle 3. From Tle 3, it cn e seen tht KSRC outperforms SRC. Especilly for Tle 1 The mximl recognition rtes (percent) of NM, NN, KERNEL-NN, SRC nd KSRC on FERET fce dtse nd the corresponding dimensions nd prmeters. Method NM NN KERNEL-NN SRC KSRC (polynomil) KSRC (Gussin) Recognition rte Dimension Prmeter d ¼ 0:8 d ¼ 2 t ¼ 2 Fig. 5. Smple imges of one right index finger on PolyU FKP dtse Recognition rte NM 0.8 NN KNN 0.78 SRC KSRC(Polynomil) KSRC(Gussin) Dimension Fig. 6. Smple imges of one plm on PolyU plmprint dtse. Fig. 8. The verge recognition rtes of NM, NN, KERNEL-NN, SRC nd KSRC versus the dimensions on ORL fce dtse cross 10 runs. Recognition rte KSRC(Polynomil) prmeter:d Recognition rte KSRC(Gussin) prmeter:t Fig. 7. The recognition rtes of KSRC on FERET fce dtse versus the vrition of the kernel prmeters: () Recognition rtes versus the prmeter d of polynomil kernel nd () recognition rtes versus the prmeter t of Gussin kernel.

7 126 J. Yin et l. / Neurocomputing 77 (2012) Tle 2 The mximl verge recognition rtes (percent) nd stndrd devitions of NM, NN, KERNEL-NN, SRC nd KSRC on ORL fce dtse cross 10 runs nd the corresponding dimensions nd prmeters. Method NM NN KERNEL-NN SRC KSRC (polynomil) KSRC (Gussin) Recognition rte 87:572:8 88:673:0 89:673:0 90:072:7 91:672:8 91:072:8 Dimension Prmeter d ¼ 0:5 d ¼ 2 t ¼ Fig. 9. Smple imge of one person under occlusion on ORL fce dtse. Tle 3 The mximl recognition rtes (percent) of SRC nd KSRC under occlusion on ORL fce dtse nd the corresponding dimensions nd prmeters. Method SRC KSRC (polynomil) KSRC (Gussin) Recognition rte NM NN 0.86 KNN SRC KSRC(Polynomil) 0.84 KSRC(Gussin) Trining smple size Recognition rte Dimension Prmeter d ¼ 2 t ¼ 3 Fig. 10. The mximl verge recognition rtes of NM, NN, KERNEL-NN, SRC nd KSRC versus the vrition of the trining smple size on Yle fce dtse. KSRC with polynomil kernel, its recognition rte is 7.5 percent more thn SRC. Therefore, KSRC is good clssifier to hndle occlusion. On Yle fce dtse, l imges per individul (l vries from 3 to 5) re rndomly selected for trining nd the remining 11 l imges re used for testing. We run the system 20 times. The first 10 times re used for prmeter selection nd the rest 10 times re used for performnce evlution of NM, NN, KERNEL-NN, SRC nd KSRC. Here RP is used for feture extrction. For RP, smples re projected into lower dimensionl feture spce using rndom mtrix whose column hs unit lengths. It hs een found to e sufficiently ccurte method for extrcting feture of high dimensionl dt. The optiml kernel prmeters re required y glol-to-locl serch strtegy. The optiml d of polynomil kernel is set s 5 for KSRC nd 0.3 for KERNEL-NN, respectively. The optiml t of Gussin kernel for KSRC is set s 1. Fig. 10 illustrtes the mximl verge recognition rtes of five methods versus the vrition of trining smple size. Fig. 10 shows tht KSRC still performs est irrespective of the vrition of trining smple size, whether polynomil kernel is used or Gussin kernel is used. SRC performs second est. Moreover, KERNEL-NN outperforms NN nd NM. These re ll consistent with the experiments on FERET nd ORL fce dtses. However, there is lso one inconsistent point tht NM performs etter thn NN in this experiment. For AR fce dtse, the first seven imges, which were tken in the first session, re used for trining while the rest seven tken in the second session re used for testing. For the PolyU FKP dtse, we use the first six FKP imges collected in the first session for trining nd the rest six collected in the second session for testing. For the PolyU plmprint dtse, the first three plmprint imges cptured in the first session re chosen for Tle 4 The mximl recognition rtes (percent) of NM, NN, KERNEL-NN, SRC nd KSRC on AR fce, the PolyU FKP nd the PolyU plmprint dtses nd the corresponding prmeters (in prentheses). trining nd the remining three cptured in the second session for testing. PCA is first performed for feture extrction nd dimension reduction. Then the dimension reduced smples re clssified y NM, NN, KERNEL-NN, SRC nd KSRC seprtely. Tle 4 lists the mximl recognition rtes nd the corresponding dimensions of ech clssifiction method on three dtses. From Tle 4, we cn see tht SRC nd KSRC still outperform other three clssifiers nd KSRC performs etter thn SRC. This demonstrtes tht KSRC is more effective thn SRC for clssifiction gin. 4. Conclusions NM NN KERNEL-NN SRC KSRC KSRC (polynomil) (Gussin) AR (d ¼ 0:5) PolyU FKP (d ¼ 0:5) PolyU plmprint (d ¼ 0:8) (d ¼ 2) 73.7 (t ¼ 5) (d ¼ 1) 71.8 (t ¼ 9) (d ¼ 1) 96.3 (t ¼ 3) SRC pplies sprse representtion coefficient to clssifiction. Sprse representtion coefficient contins very importnt discriminting informtion, so SRC hs more powerful discriminting ility thn clssifiction methods such s NM, NN nd KERNEL-NN. In this pper, we develop kernel sprse representtion sed clssifiction (KSRC) lgorithm. For KSRC, smples re mpped from originl

8 J. Yin et l. / Neurocomputing 77 (2012) feture spce into high dimensionl feture spce first, nd then SRC is performed in the high dimensionl feture spce. Although the explicit smples in the high dimensionl feture spce re unknown, we prove tht SRC could e implemented successfully using kernel function. If n pproprite kernel is utilized, sprse representtion coefficient of the test smple in the high dimensionl feture spce will reflect its lel informtion more ccurtely. Nmely, sprse representtion coefficient in the high dimensionl feture spce contins more effective discriminting informtion thn sprse representtion coefficient in the originl feture spce. Hence, KSRC could otin higher recognition rte thn SRC. Experimentl results on FERET, ORL, Yle nd AR fce dtses nd the PolyU plmprint nd FKP dtses indicte the effectiveness of KSRC. For smples contining smll noise, SRC nd KSRC hve the sme computtionl cost. For smples contining ig noise, if the numer of trining smple size is smller thn the dimension of smple, KSRC is more efficient thn SRC. Contrrily, SRC is more efficient thn KSRC. Besides, when performing KSRC, we should find the optiml prmeters. This process will increse its computtionl cost. Acknowledgments This work is supported y the Ntionl Science Foundtion of Chin under Grnt nos , , nd Appendix A The Proof of Theorem 1. Let GðÞ¼ðB fðyþþ T ðb fðyþþ The derivtive of GðÞ with respect to the vrile is G 0 ðþ¼2b T B 2B T fðyþ Since GðÞZ0, GðÞ chieves the minimum vlue when B T B B T fðyþ¼0 If B T B B T fðyþ is close to 0, GðÞ will pproch its minimum vlue nd qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi p :B fðyþ: 2 ¼ ðb fðyþþ T ðb fðyþþ ¼ GðÞ ffiffiffiffiffiffiffiffiffiffi Society Workshop Neurl Networks for Signl Processing, vol. IX, 1999, pp [6] S. Mike, G. Rätsch, B. Schölkopf, A. Smol, J. Weston, K.R. Müller, Invrint feture extrction nd clssifiction in kernel spces, in: Proceedings of the 13th Annul Neurl Informtion Processing Systems Conference, 1999, pp [7] K. Yu, L. Ji, X.G. Zhng, Kernel nerest-neighor lgorithm, Neurl Process. Lett. 15 (2002) [8] J. Yng, J. Wright, T. Hung, Y. M, Imge super-resolution s sprse representtion of rw ptches, in: Proceedings of the IEEE Interntionl Conference on Computer Vision nd Pttern Recognition, [9] S. Ro, R. Tron, R. Vidl, Y. M, Motion segmenttion vi roust suspce seprtion in the presence of outlying, incomplete, nd corrupted trjectories, in: Proceedings of the IEEE Interntionl Conference on Computer Vision nd Pttern Recognition, [10] J. Mirl, G. Spiro, M. Eld, Lerning multiscle sprse representtions for imge nd video restortion, SIAM MMS 7 (1) (2008) [11] J. Wright, A.Y. Yng, A. Gnesh, S.S. Sstry, Y. M, Roust fce recognition vi sprse representtion, IEEE Trns. Pttern Anl. Mch. Intell. 31 (2) (2009) [12] E. Amldi, V. Knn, On the pproximility of minimizing nonzero vriles or unstisfied reltions in liner systems, Theor. Comput. Sci. 209 (1998) [13] E. Cnde, J. Romerg, T. To, Stle signl recovery from incomplete nd inccurte mesurements, Commun. Pure Appl. Mth. 59 (8) (2006) [14] D. Donoho, For most lrge underdetermined systems of liner equtions the miniml l 1 -norm solution is lso the sprest solution, Commun. Pure Appl. Mth. 59 (6) (2006) [15] E. Cnde, T. To, Nerest-optiml signl recovery from rndom projections: universl encoding strtegies, IEEE Trns. Inf. Theory 52 (12) (2006) [16] S. Chen, D. Donoho, M. Sunders, Atomic decomposition y sis pursuit, SIAM Rev. 43 (1) (2001) [17] M. Turk, A.P. Pentlnd, Fce recognition using eigenfces, in: Proceeding of the IEEE interntionl Conference on Computer Vision nd Pttern Recognition, 1991, pp [18] E. Binhm, H. Mnnil, Rndom projection in dimension reduction: ppliction to imge nd text dt, in: Proceeding of the Seventh ACM SIGKDD Interntionl Conference on Knowledge Discovery nd Dt Mining, 2001, pp [19] P.J. Phillips, H. Moon, S.A. Rizvi, P.,.J. Russ, The FERET evlution methodology for fce-recognition lgorithms, IEEE Trns. Pttern Anl. Mch. Intell. 22 (10) (2000) [20] A. Mrtinez, R. Benvente, The AR fce dtse, CVC Tech. Rep. 24 (1998). [21] L. Zhng, L. Zhng, D. Zhng, H.L. Zhu, On-line finger-knuckle-print verifiction for personl uthentiction, Pttern Recognition 43 (7) (2010) [22] D. Zhng, Plmprint Authentiction, Kluwer Acdemic, [23] D. Donoho, Y. Tsig, Fst solution of l 1 -norm minimiztion prolems when the solution my e sprse, Preprint, / reserch.htmls, Jun Yin received BS degree in Mthemtics nd PhD degree in Pttern Recognition nd Intelligence System from Nnjing University of Science nd Technology in 2006 nd 2011, respectively. His current reserch interest includes pttern recognition, fce recognition nd mchine lerning. will lso pproch its minimum vlue. Therefore, for ny ez0, there must exist dz0 such tht if :B T B B T fðyþ: 2 rd :B fðyþ: 2 re will e stisfied. & References [1] A.K. Jin, P.W. Duin, J.C. Mo, Sttisticl pttern recogniton: review, IEEE Trns. Pttern Anl. Mch. Intell. 22 (1) (2000) [2] S.B. Kotsintis, Supervised mchine lerning: review of clssifiction techniques, Informtic 31 (2007) [3] K.R. Müller, S. Mike, G. Rätsch, K. Tsud, B. Schölkopf, An introduction to kernel-sed lerning lgorithms, IEEE Trns. Neurl Networks 12 (2) (2001) [4] B. Schölkopf, A. Smol, K.R. Müller, Nonliner component nlysis s kernel eigenvlue prolem, Neurl Comput. 10 (5) (1998) [5] S. Mike, G. Rätsch, J. Weston, B. Schölkopf, K.R. Müller, Fisher discriminnt nlysis with kernels, in: Proceedings of the 1999 IEEE Signl Processing Zhonghu Liu received BS degree in Computer Engineering from the First Aeronuticl Institute of the Air Force, MS degree in Computer Softwre nd Theory from Xihu University nd PhD degree in Pttern Recognition nd Intelligence System from Nnjing University of Science nd Technology in 1998, 2005 nd 2011, respectively. His current reserch interest includes pttern recognition, fce recognition nd imge processing.

9 128 J. Yin et l. / Neurocomputing 77 (2012) Zhong Jin received BS degree in Mthemtics, MS degree in Applied Mthemtics nd PhD degree in Pttern Recognition nd Intelligence System from Nnjing University of Science nd Technology in 1982, 1984 nd 1999, respectively. Now he is professor in the Deprtment of Computer Science nd Technology t Nnjing University of Science nd Technology. His current interest includes pttern recognition, imge processing nd fce recognition. Wnkou Yng received BS degree in Computer Science nd Technology, MS degree nd PhD degree in Pttern Recognition nd Intelligence System from Nnjing University of Science nd Technology in 2002, 2004 nd 2009, respectively. Now he is Postdoctorl Fellow in the school of utomtion t Southest University. His reserch interest includes pttern recognition, computer vision nd digitl imge processing.

Lecture 10 Evolutionary Computation: Evolution strategies and genetic programming

Lecture 10 Evolutionary Computation: Evolution strategies and genetic programming Lecture 10 Evolutionry Computtion: Evolution strtegies nd genetic progrmming Evolution strtegies Genetic progrmming Summry Negnevitsky, Person Eduction, 2011 1 Evolution Strtegies Another pproch to simulting

More information

Before We Begin. Introduction to Spatial Domain Filtering. Introduction to Digital Image Processing. Overview (1): Administrative Details (1):

Before We Begin. Introduction to Spatial Domain Filtering. Introduction to Digital Image Processing. Overview (1): Administrative Details (1): Overview (): Before We Begin Administrtive detils Review some questions to consider Winter 2006 Imge Enhncement in the Sptil Domin: Bsics of Sptil Filtering, Smoothing Sptil Filters, Order Sttistics Filters

More information

Complete Coverage Path Planning of Mobile Robot Based on Dynamic Programming Algorithm Peng Zhou, Zhong-min Wang, Zhen-nan Li, Yang Li

Complete Coverage Path Planning of Mobile Robot Based on Dynamic Programming Algorithm Peng Zhou, Zhong-min Wang, Zhen-nan Li, Yang Li 2nd Interntionl Conference on Electronic & Mechnicl Engineering nd Informtion Technology (EMEIT-212) Complete Coverge Pth Plnning of Mobile Robot Bsed on Dynmic Progrmming Algorithm Peng Zhou, Zhong-min

More information

A Tautology Checker loosely related to Stålmarck s Algorithm by Martin Richards

A Tautology Checker loosely related to Stålmarck s Algorithm by Martin Richards A Tutology Checker loosely relted to Stålmrck s Algorithm y Mrtin Richrds mr@cl.cm.c.uk http://www.cl.cm.c.uk/users/mr/ University Computer Lortory New Museum Site Pemroke Street Cmridge, CB2 3QG Mrtin

More information

Slides for Data Mining by I. H. Witten and E. Frank

Slides for Data Mining by I. H. Witten and E. Frank Slides for Dt Mining y I. H. Witten nd E. Frnk Simplicity first Simple lgorithms often work very well! There re mny kinds of simple structure, eg: One ttriute does ll the work All ttriutes contriute eqully

More information

Machine vision system for surface inspection on brushed industrial parts.

Machine vision system for surface inspection on brushed industrial parts. Mchine vision system for surfce inspection on rushed industril prts. Nicols Bonnot, Rlph Seulin, Frederic Merienne Lortoire Le2i, CNRS UMR 5158, University of Burgundy, Le Creusot, Frnce. ABSTRACT This

More information

Statistical classification of spatial relationships among mathematical symbols

Statistical classification of spatial relationships among mathematical symbols 2009 10th Interntionl Conference on Document Anlysis nd Recognition Sttisticl clssifiction of sptil reltionships mong mthemticl symbols Wl Aly, Seiichi Uchid Deprtment of Intelligent Systems, Kyushu University

More information

P(r)dr = probability of generating a random number in the interval dr near r. For this probability idea to make sense we must have

P(r)dr = probability of generating a random number in the interval dr near r. For this probability idea to make sense we must have Rndom Numers nd Monte Crlo Methods Rndom Numer Methods The integrtion methods discussed so fr ll re sed upon mking polynomil pproximtions to the integrnd. Another clss of numericl methods relies upon using

More information

GENERATING ORTHOIMAGES FOR CLOSE-RANGE OBJECTS BY AUTOMATICALLY DETECTING BREAKLINES

GENERATING ORTHOIMAGES FOR CLOSE-RANGE OBJECTS BY AUTOMATICALLY DETECTING BREAKLINES GENEATING OTHOIMAGES FO CLOSE-ANGE OBJECTS BY AUTOMATICALLY DETECTING BEAKLINES Efstrtios Stylinidis 1, Lzros Sechidis 1, Petros Ptis 1, Spiros Sptls 2 Aristotle University of Thessloniki 1 Deprtment of

More information

L. Yaroslavsky. Fundamentals of Digital Image Processing. Course

L. Yaroslavsky. Fundamentals of Digital Image Processing. Course L. Yroslvsky. Fundmentls of Digitl Imge Processing. Course 0555.330 Lecture. Imge enhncement.. Imge enhncement s n imge processing tsk. Clssifiction of imge enhncement methods Imge enhncement is processing

More information

2 Computing all Intersections of a Set of Segments Line Segment Intersection

2 Computing all Intersections of a Set of Segments Line Segment Intersection 15-451/651: Design & Anlysis of Algorithms Novemer 14, 2016 Lecture #21 Sweep-Line nd Segment Intersection lst chnged: Novemer 8, 2017 1 Preliminries The sweep-line prdigm is very powerful lgorithmic design

More information

Agilent Mass Hunter Software

Agilent Mass Hunter Software Agilent Mss Hunter Softwre Quick Strt Guide Use this guide to get strted with the Mss Hunter softwre. Wht is Mss Hunter Softwre? Mss Hunter is n integrl prt of Agilent TOF softwre (version A.02.00). Mss

More information

Text mining: bag of words representation and beyond it

Text mining: bag of words representation and beyond it Text mining: bg of words representtion nd beyond it Jsmink Dobš Fculty of Orgniztion nd Informtics University of Zgreb 1 Outline Definition of text mining Vector spce model or Bg of words representtion

More information

What are suffix trees?

What are suffix trees? Suffix Trees 1 Wht re suffix trees? Allow lgorithm designers to store very lrge mount of informtion out strings while still keeping within liner spce Allow users to serch for new strings in the originl

More information

II. THE ALGORITHM. A. Depth Map Processing

II. THE ALGORITHM. A. Depth Map Processing Lerning Plnr Geometric Scene Context Using Stereo Vision Pul G. Bumstrck, Bryn D. Brudevold, nd Pul D. Reynolds {pbumstrck,brynb,pulr2}@stnford.edu CS229 Finl Project Report December 15, 2006 Abstrct A

More information

Object and image indexing based on region connection calculus and oriented matroid theory

Object and image indexing based on region connection calculus and oriented matroid theory Discrete Applied Mthemtics 147 (2005) 345 361 www.elsevier.com/locte/dm Oject nd imge indexing sed on region connection clculus nd oriented mtroid theory Ernesto Stffetti, Antoni Gru, Frncesc Serrtos c,

More information

Efficient K-NN Search in Polyphonic Music Databases Using a Lower Bounding Mechanism

Efficient K-NN Search in Polyphonic Music Databases Using a Lower Bounding Mechanism Efficient K-NN Serch in Polyphonic Music Dtses Using Lower Bounding Mechnism Ning-Hn Liu Deprtment of Computer Science Ntionl Tsing Hu University Hsinchu,Tiwn 300, R.O.C 886-3-575679 nhliou@yhoo.com.tw

More information

On the Detection of Step Edges in Algorithms Based on Gradient Vector Analysis

On the Detection of Step Edges in Algorithms Based on Gradient Vector Analysis On the Detection of Step Edges in Algorithms Bsed on Grdient Vector Anlysis A. Lrr6, E. Montseny Computer Engineering Dept. Universitt Rovir i Virgili Crreter de Slou sin 43006 Trrgon, Spin Emil: lrre@etse.urv.es

More information

USING HOUGH TRANSFORM IN LINE EXTRACTION

USING HOUGH TRANSFORM IN LINE EXTRACTION Stylinidis, Efstrtios USING HOUGH TRANSFORM IN LINE EXTRACTION Efstrtios STYLIANIDIS, Petros PATIAS The Aristotle University of Thessloniki, Deprtment of Cdstre Photogrmmetry nd Crtogrphy Univ. Box 473,

More information

OUTPUT DELIVERY SYSTEM

OUTPUT DELIVERY SYSTEM Differences in ODS formtting for HTML with Proc Print nd Proc Report Lur L. M. Thornton, USDA-ARS, Animl Improvement Progrms Lortory, Beltsville, MD ABSTRACT While Proc Print is terrific tool for dt checking

More information

Analysis of Computed Diffraction Pattern Diagram for Measuring Yarn Twist Angle

Analysis of Computed Diffraction Pattern Diagram for Measuring Yarn Twist Angle Textiles nd Light ndustril Science nd Technology (TLST) Volume 3, 2014 DO: 10.14355/tlist.2014.0301.01 http://www.tlist-journl.org Anlysis of Computed Diffrction Pttern Digrm for Mesuring Yrn Twist Angle

More information

Ranking of Hexagonal Fuzzy Numbers for Solving Multi Objective Fuzzy Linear Programming Problem

Ranking of Hexagonal Fuzzy Numbers for Solving Multi Objective Fuzzy Linear Programming Problem Interntionl Journl of Computer pplictions 097 8887 Volume 8 No 8 Decemer 0 nking of egonl Fuzzy Numers Solving ulti Ojective Fuzzy Liner Progrmming Prolem jrjeswri. P Deprtment of themtics Chikknn Government

More information

HOPC: A NOVEL SIMILARITY METRIC BASED ON GEOMETRIC STRUCTURAL PROPERTIES FOR MULTI-MODAL REMOTE SENSING IMAGE MATCHING

HOPC: A NOVEL SIMILARITY METRIC BASED ON GEOMETRIC STRUCTURAL PROPERTIES FOR MULTI-MODAL REMOTE SENSING IMAGE MATCHING ISPRS Annls of the Photogrmmetry, Remote Sensing nd Sptil Informtion Sciences, Volume III-1, 216 XXIII ISPRS Congress, 12 19 July 216, Prgue, Czech Republic : A NOVEL SILARITY METRIC BASED ON GEOMETRIC

More information

ZZ - Advanced Math Review 2017

ZZ - Advanced Math Review 2017 ZZ - Advnced Mth Review Mtrix Multipliction Given! nd! find the sum of the elements of the product BA First, rewrite the mtrices in the correct order to multiply The product is BA hs order x since B is

More information

In the last lecture, we discussed how valid tokens may be specified by regular expressions.

In the last lecture, we discussed how valid tokens may be specified by regular expressions. LECTURE 5 Scnning SYNTAX ANALYSIS We know from our previous lectures tht the process of verifying the syntx of the progrm is performed in two stges: Scnning: Identifying nd verifying tokens in progrm.

More information

Image Segmentation Using Wavelet and watershed transform

Image Segmentation Using Wavelet and watershed transform Imge Segmenttion Using Wvelet nd wtershed trnsform Atollh Hdddi, Mhmod R. Shei, Mohmmd J. Vldn Zoej, Ali mohmmdzdeh Fculty of Geodesy nd Geomtics Engineering, K. N. Toosi University of Technology, Vli_Asr

More information

Subband coding of image sequences using multiple vector quantizers. Emanuel Martins, Vitor Silva and Luís de Sá

Subband coding of image sequences using multiple vector quantizers. Emanuel Martins, Vitor Silva and Luís de Sá Sund coding of imge sequences using multiple vector quntizers Emnuel Mrtins, Vitor Silv nd Luís de Sá Instituto de Telecomunicções, Deprtmento de Engenhri Electrotécnic Pólo II d Universidde de Coimr,

More information

Product of polynomials. Introduction to Programming (in C++) Numerical algorithms. Product of polynomials. Product of polynomials

Product of polynomials. Introduction to Programming (in C++) Numerical algorithms. Product of polynomials. Product of polynomials Product of polynomils Introduction to Progrmming (in C++) Numericl lgorithms Jordi Cortdell, Ricrd Gvldà, Fernndo Orejs Dept. of Computer Science, UPC Given two polynomils on one vrile nd rel coefficients,

More information

Outline. Two combinatorial optimization problems in machine learning. Talk objectives. Grammar induction. DFA induction.

Outline. Two combinatorial optimization problems in machine learning. Talk objectives. Grammar induction. DFA induction. Outline Two comintoril optimiztion prolems in mchine lerning Pierre.Dupont@uclouvin.e 1 Feture selection ICTEAM Institute Université ctholique de Louvin Belgium My 1, 011 P. Dupont (UCL Mchine Lerning

More information

Memory-Optimized Software Synthesis from Dataflow Program Graphs withlargesizedatasamples

Memory-Optimized Software Synthesis from Dataflow Program Graphs withlargesizedatasamples EURSIP Journl on pplied Signl Processing 2003:6, 54 529 c 2003 Hindwi Publishing orportion Memory-Optimized Softwre Synthesis from tflow Progrm Grphs withlrgesizetsmples Hyunok Oh The School of Electricl

More information

Cone Cluster Labeling for Support Vector Clustering

Cone Cluster Labeling for Support Vector Clustering Cone Cluster Lbeling for Support Vector Clustering Sei-Hyung Lee Deprtment of Computer Science University of Msschusetts Lowell MA 1854, U.S.A. slee@cs.uml.edu Kren M. Dniels Deprtment of Computer Science

More information

Lily Yen and Mogens Hansen

Lily Yen and Mogens Hansen SKOLID / SKOLID No. 8 Lily Yen nd Mogens Hnsen Skolid hs joined Mthemticl Myhem which is eing reformtted s stnd-lone mthemtics journl for high school students. Solutions to prolems tht ppered in the lst

More information

COMP 423 lecture 11 Jan. 28, 2008

COMP 423 lecture 11 Jan. 28, 2008 COMP 423 lecture 11 Jn. 28, 2008 Up to now, we hve looked t how some symols in n lphet occur more frequently thn others nd how we cn sve its y using code such tht the codewords for more frequently occuring

More information

If you are at the university, either physically or via the VPN, you can download the chapters of this book as PDFs.

If you are at the university, either physically or via the VPN, you can download the chapters of this book as PDFs. Lecture 5 Wlks, Trils, Pths nd Connectedness Reding: Some of the mteril in this lecture comes from Section 1.2 of Dieter Jungnickel (2008), Grphs, Networks nd Algorithms, 3rd edition, which is ville online

More information

IMAGE QUALITY OPTIMIZATION BASED ON WAVELET FILTER DESIGN AND WAVELET DECOMPOSITION IN JPEG2000. Do Quan and Yo-Sung Ho

IMAGE QUALITY OPTIMIZATION BASED ON WAVELET FILTER DESIGN AND WAVELET DECOMPOSITION IN JPEG2000. Do Quan and Yo-Sung Ho IMAGE QUALITY OPTIMIZATIO BASED O WAVELET FILTER DESIG AD WAVELET DECOMPOSITIO I JPEG2000 Do Qun nd Yo-Sung Ho School of Informtion & Mechtronics Gwngju Institute of Science nd Technology (GIST) 26 Cheomdn-gwgiro

More information

Approximation by NURBS with free knots

Approximation by NURBS with free knots pproximtion by NURBS with free knots M Rndrinrivony G Brunnett echnicl University of Chemnitz Fculty of Computer Science Computer Grphics nd Visuliztion Strße der Ntionen 6 97 Chemnitz Germny Emil: mhrvo@informtiktu-chemnitzde

More information

A dual of the rectangle-segmentation problem for binary matrices

A dual of the rectangle-segmentation problem for binary matrices A dul of the rectngle-segmenttion prolem for inry mtrices Thoms Klinowski Astrct We consider the prolem to decompose inry mtrix into smll numer of inry mtrices whose -entries form rectngle. We show tht

More information

Performance enhancement of IEEE DCF using novel backoff algorithm

Performance enhancement of IEEE DCF using novel backoff algorithm Kuo et l. EURASIP Journl on Wireless Communictions nd Networking 212, 212:274 http://jis.eursipjournls.com/content/212/1/274 RESEARCH Open Access Performnce enhncement of IEEE 82.11 using novel ckoff lgorithm

More information

A Transportation Problem Analysed by a New Ranking Method

A Transportation Problem Analysed by a New Ranking Method (IJIRSE) Interntionl Journl of Innovtive Reserch in Science & Engineering ISSN (Online) 7-07 A Trnsporttion Problem Anlysed by New Rnking Method Dr. A. Shy Sudh P. Chinthiy Associte Professor PG Scholr

More information

CHAPTER III IMAGE DEWARPING (CALIBRATION) PROCEDURE

CHAPTER III IMAGE DEWARPING (CALIBRATION) PROCEDURE CHAPTER III IMAGE DEWARPING (CALIBRATION) PROCEDURE 3.1 Scheimpflug Configurtion nd Perspective Distortion Scheimpflug criterion were found out to be the best lyout configurtion for Stereoscopic PIV, becuse

More information

documents 1. Introduction

documents 1. Introduction www.ijcsi.org 4 Efficient structurl similrity computtion etween XML documents Ali Aïtelhdj Computer Science Deprtment, Fculty of Electricl Engineering nd Computer Science Mouloud Mmmeri University of Tizi-Ouzou

More information

An Algorithm for Enumerating All Maximal Tree Patterns Without Duplication Using Succinct Data Structure

An Algorithm for Enumerating All Maximal Tree Patterns Without Duplication Using Succinct Data Structure , Mrch 12-14, 2014, Hong Kong An Algorithm for Enumerting All Mximl Tree Ptterns Without Dupliction Using Succinct Dt Structure Yuko ITOKAWA, Tomoyuki UCHIDA nd Motoki SANO Astrct In order to extrct structured

More information

Inference of node replacement graph grammars

Inference of node replacement graph grammars Glley Proof 22/6/27; :6 File: id293.tex; BOKCTP/Hin p. Intelligent Dt Anlysis (27) 24 IOS Press Inference of node replcement grph grmmrs Jcek P. Kukluk, Lwrence B. Holder nd Dine J. Cook Deprtment of Computer

More information

Presentation Martin Randers

Presentation Martin Randers Presenttion Mrtin Rnders Outline Introduction Algorithms Implementtion nd experiments Memory consumption Summry Introduction Introduction Evolution of species cn e modelled in trees Trees consist of nodes

More information

Algorithm Design (5) Text Search

Algorithm Design (5) Text Search Algorithm Design (5) Text Serch Tkshi Chikym School of Engineering The University of Tokyo Text Serch Find sustring tht mtches the given key string in text dt of lrge mount Key string: chr x[m] Text Dt:

More information

A Fast Imaging Algorithm for Near Field SAR

A Fast Imaging Algorithm for Near Field SAR Journl of Computing nd Electronic Informtion Mngement ISSN: 2413-1660 A Fst Imging Algorithm for Ner Field SAR Guoping Chen, Lin Zhng, * College of Optoelectronic Engineering, Chongqing University of Posts

More information

Compression Outline :Algorithms in the Real World. Lempel-Ziv Algorithms. LZ77: Sliding Window Lempel-Ziv

Compression Outline :Algorithms in the Real World. Lempel-Ziv Algorithms. LZ77: Sliding Window Lempel-Ziv Compression Outline 15-853:Algorithms in the Rel World Dt Compression III Introduction: Lossy vs. Lossless, Benchmrks, Informtion Theory: Entropy, etc. Proility Coding: Huffmn + Arithmetic Coding Applictions

More information

A Comparison of the Discretization Approach for CST and Discretization Approach for VDM

A Comparison of the Discretization Approach for CST and Discretization Approach for VDM Interntionl Journl of Innovtive Reserch in Advnced Engineering (IJIRAE) Volume1 Issue1 (Mrch 2014) A Comprison of the Discretiztion Approch for CST nd Discretiztion Approch for VDM Omr A. A. Shib Fculty

More information

A New Learning Algorithm for the MAXQ Hierarchical Reinforcement Learning Method

A New Learning Algorithm for the MAXQ Hierarchical Reinforcement Learning Method A New Lerning Algorithm for the MAXQ Hierrchicl Reinforcement Lerning Method Frzneh Mirzzdeh 1, Bbk Behsz 2, nd Hmid Beigy 1 1 Deprtment of Computer Engineering, Shrif University of Technology, Tehrn,

More information

Linear analysis of texture property relationships using process-based representations of Rodrigues space

Linear analysis of texture property relationships using process-based representations of Rodrigues space Act Mterili 55 (7) 573 587 www.ctmt-journls.com Liner nlysis of texture property reltionships using process-sed representtions of Rodrigues spce Veer Sundrrghvn, Nichols Zrs * Mterils Process Design nd

More information

Spectral Analysis of MCDF Operations in Image Processing

Spectral Analysis of MCDF Operations in Image Processing Spectrl Anlysis of MCDF Opertions in Imge Processing ZHIQIANG MA 1,2 WANWU GUO 3 1 School of Computer Science, Northest Norml University Chngchun, Jilin, Chin 2 Deprtment of Computer Science, JilinUniversity

More information

Shape Representation and Indexing Based on Region Connection Calculus and Oriented Matroid Theory

Shape Representation and Indexing Based on Region Connection Calculus and Oriented Matroid Theory Shpe Representtion nd Indexing Bsed on Region Connection Clculus nd Oriented Mtroid Theory Ernesto Stffetti 1, Antoni Gru 2, Frncesc Serrtos 3, nd Alerto Snfeliu 1 1 Institute of Industril Rootics (CSIC-UPC)

More information

COLOUR IMAGE MATCHING FOR DTM GENERATION AND HOUSE EXTRACTION

COLOUR IMAGE MATCHING FOR DTM GENERATION AND HOUSE EXTRACTION Hee Ju Prk OLOUR IMAGE MATHING FOR DTM GENERATION AND HOUSE EXTRATION Hee Ju PARK, Petr ZINMMERMANN * Swiss Federl Institute of Technology, Zuric Switzerlnd Institute for Geodesy nd Photogrmmetry heeju@ns.shingu-c.c.kr

More information

Video-rate Image Segmentation by means of Region Splitting and Merging

Video-rate Image Segmentation by means of Region Splitting and Merging Video-rte Imge Segmenttion y mens of Region Splitting nd Merging Knur Anej, Florence Lguzet, Lionel Lcssgne, Alin Merigot Institute for Fundmentl Electronics, University of Pris South Orsy, Frnce knur.nej@gmil.com,

More information

An Effective Thresholding Technique for Otsu s Method using Contrast Enhancement 1 A Sankar Reddy, 2 C Usha Rani, 3 M. Sudhakar

An Effective Thresholding Technique for Otsu s Method using Contrast Enhancement 1 A Sankar Reddy, 2 C Usha Rani, 3 M. Sudhakar www.ijecs.in Interntionl Journl Of Engineering And Computer Science ISSN:39-74 Volume 6 Issue 8 August 07, Pge No. 96-304 Index Copernicus vlue (05): 58.0 DOI: 0.8535/ijecs/v6i8. An Effective Thresholding

More information

Out of Focus Blur Estimation Using Genetic Algorithm

Out of Focus Blur Estimation Using Genetic Algorithm Journl of Computer Science 4 (4): 98-304, 008 ISSN 1549-3636 008 Science Pulictions Out of Focus Blur Estimtion Using Genetic Algorithm Mohsen Erhimi Moghddm Deprtment of Electricl nd Computer Engineering,

More information

Expert Systems with Applications

Expert Systems with Applications Expert Systems with Applictions 36 (2009) 2023 2035 Contents lists ville t ScienceDirect Expert Systems with Applictions journl homepge: www.elsevier.com/locte/esw A new document representtion using term

More information

9 Graph Cutting Procedures

9 Graph Cutting Procedures 9 Grph Cutting Procedures Lst clss we begn looking t how to embed rbitrry metrics into distributions of trees, nd proved the following theorem due to Brtl (1996): Theorem 9.1 (Brtl (1996)) Given metric

More information

Distributed Systems Principles and Paradigms

Distributed Systems Principles and Paradigms Distriuted Systems Principles nd Prdigms Chpter 11 (version April 7, 2008) Mrten vn Steen Vrije Universiteit Amsterdm, Fculty of Science Dept. Mthemtics nd Computer Science Room R4.20. Tel: (020) 598 7784

More information

LETKF compared to 4DVAR for assimilation of surface pressure observations in IFS

LETKF compared to 4DVAR for assimilation of surface pressure observations in IFS LETKF compred to 4DVAR for ssimiltion of surfce pressure oservtions in IFS Pu Escrià, Mssimo Bonvit, Mts Hmrud, Lrs Isksen nd Pul Poli Interntionl Conference on Ensemle Methods in Geophysicl Sciences Toulouse,

More information

CS321 Languages and Compiler Design I. Winter 2012 Lecture 5

CS321 Languages and Compiler Design I. Winter 2012 Lecture 5 CS321 Lnguges nd Compiler Design I Winter 2012 Lecture 5 1 FINITE AUTOMATA A non-deterministic finite utomton (NFA) consists of: An input lphet Σ, e.g. Σ =,. A set of sttes S, e.g. S = {1, 3, 5, 7, 11,

More information

A Sparse Grid Representation for Dynamic Three-Dimensional Worlds

A Sparse Grid Representation for Dynamic Three-Dimensional Worlds A Sprse Grid Representtion for Dynmic Three-Dimensionl Worlds Nthn R. Sturtevnt Deprtment of Computer Science University of Denver Denver, CO, 80208 sturtevnt@cs.du.edu Astrct Grid representtions offer

More information

The Greedy Method. The Greedy Method

The Greedy Method. The Greedy Method Lists nd Itertors /8/26 Presenttion for use with the textook, Algorithm Design nd Applictions, y M. T. Goodrich nd R. Tmssi, Wiley, 25 The Greedy Method The Greedy Method The greedy method is generl lgorithm

More information

ARTIFICIAL NEURAL NETWORK FOR TEXTURE CLASSIFICATION USING SEVERAL FEATURES: A COMPARATIVE STUDY

ARTIFICIAL NEURAL NETWORK FOR TEXTURE CLASSIFICATION USING SEVERAL FEATURES: A COMPARATIVE STUDY ATIFICIAL NEUAL NETWOK FO TEXTUE CLASSIFICATION USING SEVEAL FEATUES: A COMPAATIVE STUDY Mohmmed W. Ashour, Khled M. Mhr, nd Mhmoud F. Hussin College of Engineering & Technology, Arb Acdemy for Science

More information

Fig.25: the Role of LEX

Fig.25: the Role of LEX The Lnguge for Specifying Lexicl Anlyzer We shll now study how to uild lexicl nlyzer from specifiction of tokens in the form of list of regulr expressions The discussion centers round the design of n existing

More information

a(e, x) = x. Diagrammatically, this is encoded as the following commutative diagrams / X

a(e, x) = x. Diagrammatically, this is encoded as the following commutative diagrams / X 4. Mon, Sept. 30 Lst time, we defined the quotient topology coming from continuous surjection q : X! Y. Recll tht q is quotient mp (nd Y hs the quotient topology) if V Y is open precisely when q (V ) X

More information

DECISION LEVEL FUSION OF LIDAR DATA AND AERIAL COLOR IMAGERY BASED ON BAYESIAN THEORY FOR URBAN AREA CLASSIFICATION

DECISION LEVEL FUSION OF LIDAR DATA AND AERIAL COLOR IMAGERY BASED ON BAYESIAN THEORY FOR URBAN AREA CLASSIFICATION DECISION LEVEL FUSION OF LIDAR DATA AND AERIAL COLOR IMAGERY BASED ON BAYESIAN THEORY FOR URBAN AREA CLASSIFICATION H. Rstiveis* School of Surveying nd Geosptil Engineering, Fculty of Engineering, University

More information

NOTES. Figure 1 illustrates typical hardware component connections required when using the JCM ICB Asset Ticket Generator software application.

NOTES. Figure 1 illustrates typical hardware component connections required when using the JCM ICB Asset Ticket Generator software application. ICB Asset Ticket Genertor Opertor s Guide Septemer, 2016 Septemer, 2016 NOTES Opertor s Guide ICB Asset Ticket Genertor Softwre Instlltion nd Opertion This document contins informtion for downloding, instlling,

More information

Chapter 2 Sensitivity Analysis: Differential Calculus of Models

Chapter 2 Sensitivity Analysis: Differential Calculus of Models Chpter 2 Sensitivity Anlysis: Differentil Clculus of Models Abstrct Models in remote sensing nd in science nd engineering, in generl re, essentilly, functions of discrete model input prmeters, nd/or functionls

More information

SUPPLEMENTARY INFORMATION

SUPPLEMENTARY INFORMATION Supplementry Figure y (m) x (m) prllel perpendiculr Distnce (m) Bird Stndrd devition for distnce (m) c 6 prllel perpendiculr 4 doi:.8/nture99 SUPPLEMENTARY FIGURE Confirmtion tht movement within the flock

More information

4452 Mathematical Modeling Lecture 4: Lagrange Multipliers

4452 Mathematical Modeling Lecture 4: Lagrange Multipliers Mth Modeling Lecture 4: Lgrnge Multipliers Pge 4452 Mthemticl Modeling Lecture 4: Lgrnge Multipliers Lgrnge multipliers re high powered mthemticl technique to find the mximum nd minimum of multidimensionl

More information

Topological Queries on Graph-structured XML Data: Models and Implementations

Topological Queries on Graph-structured XML Data: Models and Implementations Topologicl Queries on Grph-structured XML Dt: Models nd Implementtions Hongzhi Wng, Jinzhong Li, nd Jizhou Luo Astrct In mny pplictions, dt is in grph structure, which cn e nturlly represented s grph-structured

More information

TOWARDS GRADIENT BASED AERODYNAMIC OPTIMIZATION OF WIND TURBINE BLADES USING OVERSET GRIDS

TOWARDS GRADIENT BASED AERODYNAMIC OPTIMIZATION OF WIND TURBINE BLADES USING OVERSET GRIDS TOWARDS GRADIENT BASED AERODYNAMIC OPTIMIZATION OF WIND TURBINE BLADES USING OVERSET GRIDS S. H. Jongsm E. T. A. vn de Weide H. W. M. Hoeijmkers Overset symposium 10-18-2012 Deprtment of mechnicl engineering

More information

Computer Vision and Image Understanding

Computer Vision and Image Understanding Computer Vision nd Imge Understnding 116 (2012) 25 37 Contents lists ville t SciVerse ScienceDirect Computer Vision nd Imge Understnding journl homepge: www.elsevier.com/locte/cviu A systemtic pproch for

More information

Computing offsets of freeform curves using quadratic trigonometric splines

Computing offsets of freeform curves using quadratic trigonometric splines Computing offsets of freeform curves using qudrtic trigonometric splines JIULONG GU, JAE-DEUK YUN, YOONG-HO JUNG*, TAE-GYEONG KIM,JEONG-WOON LEE, BONG-JUN KIM School of Mechnicl Engineering Pusn Ntionl

More information

Some necessary and sufficient conditions for two variable orthogonal designs in order 44

Some necessary and sufficient conditions for two variable orthogonal designs in order 44 University of Wollongong Reserch Online Fculty of Informtics - Ppers (Archive) Fculty of Engineering n Informtion Sciences 1998 Some necessry n sufficient conitions for two vrile orthogonl esigns in orer

More information

Taming Subgraph Isomorphism for RDF Query Processing

Taming Subgraph Isomorphism for RDF Query Processing Tming Sugrph Isomorphism for RDF Query Processing Jinh Kim # jinh.kim@orcle.com Hyungyu Shin hgshin@dl.postech.c.kr Sungpck Hong # Hssn Chfi # {sungpck.hong, hssn.chfi}@orcle.com POSTECH, South Kore #

More information

Towards the rectification of highly distorted texts

Towards the rectification of highly distorted texts Towrds the rectifiction of highly distorted texts Ştefni Cǎlǎrǎşnu 1, Séverine Duuisson 2 nd Jonthn Frizio 1 1 EPITA-LRDE, 14-16, rue Voltire, F-94276, Le Kremlin Bicêtre, Frnce 2 Soronne Universités,

More information

F. R. K. Chung y. University ofpennsylvania. Philadelphia, Pennsylvania R. L. Graham. AT&T Labs - Research. March 2,1997.

F. R. K. Chung y. University ofpennsylvania. Philadelphia, Pennsylvania R. L. Graham. AT&T Labs - Research. March 2,1997. Forced convex n-gons in the plne F. R. K. Chung y University ofpennsylvni Phildelphi, Pennsylvni 19104 R. L. Grhm AT&T Ls - Reserch Murry Hill, New Jersey 07974 Mrch 2,1997 Astrct In seminl pper from 1935,

More information

Unit #9 : Definite Integral Properties, Fundamental Theorem of Calculus

Unit #9 : Definite Integral Properties, Fundamental Theorem of Calculus Unit #9 : Definite Integrl Properties, Fundmentl Theorem of Clculus Gols: Identify properties of definite integrls Define odd nd even functions, nd reltionship to integrl vlues Introduce the Fundmentl

More information

SAPPER: Subgraph Indexing and Approximate Matching in Large Graphs

SAPPER: Subgraph Indexing and Approximate Matching in Large Graphs SAPPER: Sugrph Indexing nd Approximte Mtching in Lrge Grphs Shijie Zhng, Jiong Yng, Wei Jin EECS Dept., Cse Western Reserve University, {shijie.zhng, jiong.yng, wei.jin}@cse.edu ABSTRACT With the emergence

More information

Character-Stroke Detection for Text-Localization and Extraction

Character-Stroke Detection for Text-Localization and Extraction Chrcter-Stroke Detection for Text-Locliztion nd Extrction Krishn Subrmnin ksubrm@bbn.com Prem Ntrjn pntrj@bbn.com Michel Decerbo mdecerbo@bbn.com Dvid Cstñòn Boston University dc@bu.edu Abstrct In this

More information

5 Regular 4-Sided Composition

5 Regular 4-Sided Composition Xilinx-Lv User Guide 5 Regulr 4-Sided Composition This tutoril shows how regulr circuits with 4-sided elements cn be described in Lv. The type of regulr circuits tht re discussed in this tutoril re those

More information

Evolutionary Approaches To Minimizing Network Coding Resources

Evolutionary Approaches To Minimizing Network Coding Resources This full text pper ws peer reviewed t the direction of IEEE Communictions Society suject mtter experts for puliction in the IEEE INFOCOM 2007 proceedings. Evolutionry Approches To Minimizing Network Coding

More information

Categorical Skylines for Streaming Data

Categorical Skylines for Streaming Data Ctegoricl Skylines for Streming Dt ABSTRACT Nikos Srks University of Toronto nsrks@cs.toronto.edu Nick Kouds University of Toronto kouds@cs.toronto.edu The prolem of skyline computtion hs ttrcted considerle

More information

Optimization of Air Bearing Slider Design

Optimization of Air Bearing Slider Design Proceedings of TC2005 orld Tribology Congress III Proceedings of TC2005 September 2-6, orld 2005, Tribology shington, Congress D.C., III SA September 2-6, 2005, shington, D.C., SA Optimiztion of Air Bering

More information

The analysis of image feature robustness using cometcloud

The analysis of image feature robustness using cometcloud J Pthol Inform Editor-in-Chief: Anil V. Prwni, Liron Pntnowitz, Pittsurgh, PA, USA Pittsurgh, PA, USA OPEN ACCESS HTML formt For entire Editoril Bord visit : www.jpthinformtics.org/editorilord.sp Reserch

More information

Revisiting the notion of Origin-Destination Traffic Matrix of the Hosts that are attached to a Switched Local Area Network

Revisiting the notion of Origin-Destination Traffic Matrix of the Hosts that are attached to a Switched Local Area Network Interntionl Journl of Distributed nd Prllel Systems (IJDPS) Vol., No.6, November 0 Revisiting the notion of Origin-Destintion Trffic Mtrix of the Hosts tht re ttched to Switched Locl Are Network Mondy

More information

Registering as a HPE Reseller. Quick Reference Guide for new Partners in Asia Pacific

Registering as a HPE Reseller. Quick Reference Guide for new Partners in Asia Pacific Registering s HPE Reseller Quick Reference Guide for new Prtners in Asi Pcific Registering s new Reseller prtner There re five min steps to e new Reseller prtner. Crete your Appliction Copyright 2017 Hewlett

More information

Registering as an HPE Reseller

Registering as an HPE Reseller Registering s n HPE Reseller Quick Reference Guide for new Prtners Mrch 2019 Registering s new Reseller prtner There re four min steps to register on the Prtner Redy Portl s new Reseller prtner: Appliction

More information

INVESTIGATION OF RESAMPLING EFFECTS ON IRS-1D PAN DATA

INVESTIGATION OF RESAMPLING EFFECTS ON IRS-1D PAN DATA INVESTIGATION OF RESAMPLING EFFECTS ON IRS-D PAN DATA Smpth Kumr P.,*, Onkr Dikshit nd YVS Murthy Geo-Informtics Division, Ntionl Remote Sensing Agency, Hyderd, Indi.-(smpth_k, murthy_yvs)@nrs.gov.in Deprtment

More information

UNIT 11. Query Optimization

UNIT 11. Query Optimization UNIT Query Optimiztion Contents Introduction to Query Optimiztion 2 The Optimiztion Process: An Overview 3 Optimiztion in System R 4 Optimiztion in INGRES 5 Implementing the Join Opertors Wei-Png Yng,

More information

Vulnerability Analysis of Electric Power Communication Network. Yucong Wu

Vulnerability Analysis of Electric Power Communication Network. Yucong Wu 2nd Interntionl Conference on Advnces in Mechnicl Engineering nd Industril Informtics (AMEII 2016 Vulnerbility Anlysis of Electric Power Communiction Network Yucong Wu Deprtment of Telecommunictions Engineering,

More information

Local Search Heuristics for NFA State Minimization Problem *

Local Search Heuristics for NFA State Minimization Problem * Int. J. Communictions, etwork nd System Sciences, 0, 5, 638-63 http://dx.doi.org/0.36/ijcns.0.5907 Pulished Online Septemer 0 (http://www.scirp.org/journl/ijcns) Locl Serch Heuristics for FA Stte inimiztion

More information

A multiview 3D modeling system based on stereo vision techniques

A multiview 3D modeling system based on stereo vision techniques Mchine Vision nd Applictions (2005) 16: 148 156 Digitl Oject Identifier (DOI) 10.1007/s00138-004-0165-2 Mchine Vision nd Applictions A multiview 3D modeling system sed on stereo vision techniques Soon-Yong

More information

MATH 25 CLASS 5 NOTES, SEP

MATH 25 CLASS 5 NOTES, SEP MATH 25 CLASS 5 NOTES, SEP 30 2011 Contents 1. A brief diversion: reltively prime numbers 1 2. Lest common multiples 3 3. Finding ll solutions to x + by = c 4 Quick links to definitions/theorems Euclid

More information

M-Historian and M-Trend

M-Historian and M-Trend Product Bulletin Issue Dte June 18, 2004 M-Historin nd The M-Historin mnges the collection nd rchiving of trend dt, nd enles the presenttion of rchived trend dt in the ssocited softwre component. M-Historin

More information

Fig.1. Let a source of monochromatic light be incident on a slit of finite width a, as shown in Fig. 1.

Fig.1. Let a source of monochromatic light be incident on a slit of finite width a, as shown in Fig. 1. Answer on Question #5692, Physics, Optics Stte slient fetures of single slit Frunhofer diffrction pttern. The slit is verticl nd illuminted by point source. Also, obtin n expression for intensity distribution

More information

AVolumePreservingMapfromCubetoOctahedron

AVolumePreservingMapfromCubetoOctahedron Globl Journl of Science Frontier Reserch: F Mthemtics nd Decision Sciences Volume 18 Issue 1 Version 1.0 er 018 Type: Double Blind Peer Reviewed Interntionl Reserch Journl Publisher: Globl Journls Online

More information

On String Matching in Chunked Texts

On String Matching in Chunked Texts On String Mtching in Chunked Texts Hnnu Peltol nd Jorm Trhio {hpeltol, trhio}@cs.hut.fi Deprtment of Computer Science nd Engineering Helsinki University of Technology P.O. Box 5400, FI-02015 HUT, Finlnd

More information