INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR ‘Admit Card for MID Semester Examination for the Session AUTUMN 2022-2023 mrECesRIS a ‘Senj Kem ELECTRONICS & ELECTRICAL COMMUNICATION ENGG, Course Name: “VISION AND INTELLIGENT SYSTEMS Hall: ‘NHR ‘Suraj Kumar Cie ‘Your Exam Time Schedule and Sitting Patter SI | Subno. Subject Name extmert| Date [Room] TimeSpan | sitting No. attern 1 | E0501 [MAGE AND viDEO er | someon [oman | @maneiaoam | t lmnocesswno 2 | cela [FOUNDATIONS OF LEARNING casas | RSET 5 frieoRy 3 | scsi |rarrenRecoomoN, Zaags | waz D ISNCHINE beret LigeNe i = | eosoit para STRUCTURE & aw | omer | —— REPRESENTATION = | ecensas jeoDpe sYSTENS Det suo _|_eeovawenrsoam | —— [Cori wil eld Cert, DEPT Capen a Un feeteetor ee 1 ge eae + Soseertestereais Eien beg etapa bebop omnia 4, Seavey Be de ‘Sates pm saplebey pr meninges ‘Eorogh teriiabion cat Noone vib 65, Excnage fms cha cl, 2 Inne cae taketh savers ood the Examinton Halls permite sain te ExmiotonHlia“ebjeionBle da inet peril The dase of eagles ai mater wee femora source sad oforasion wh othe iter inside rosie te Examination beg oman excangos sch sep ier aio ouside the ‘anyone adopts fae means she ese wl be frwarded to Malpractice comic. This may led oregon to nade yea suypensan and Gbaren rom schelarsipy, awards and meds as per Paes me Signature of Student in full (Geseraton Date Sep 13,202 7.01 PM, (Gcoented By: ERP, IT KGP (& scanned with OKEN Scanneroy INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR Mid-Autumn Semester Examination 2022-23 iF < pate of Examination: 28-29-2022 Session (FVAN) MB FDuration: 2hrs Full Marks: 3O Subject No.:_ EC 61203 Subject: PouNDATIONS ©F LEARNING THRORY Department/Gentensehoot E> ECE - - specific charts, graph paper, log book ete., required NA special instructions (if any): re eet Answer all questions. Make necessary assumptions if required, For N independen random variables {X[n],10 = O..Le,N J}, where (Xi) = x Oand Var(X fpf) = 1 ¥n, prove that ¥ = Je S> Xn] is # standard normal random variable. 6) ‘Suppose a biased coin with unknown P(Head) = pis tossed N times and the random variable X counts the number of heads. In this estimation problem, compute the Fisher information and the Cramer-Rao Lower Bound, Using the Cramer-Rao factorization. find out the minimum variance unbiased estimator(MVUE) of Pp is s. Consider the observation model for the unknown parameter 4: Xn] = 5sin(200en + 6) + Wn]. 1 = O.1.....N 1, where Wln} ~ (0,1) is the white gaussian noise. Find the suficient staitis forthe phase e&timator &. Find the Maximum Likelihood Estimator (MLE). Is it unbiased? Is it of minimum variance? 6) Suppose wwe have NV two dimensional data points (yfa),2{e)), m= 0,1,-...07~ 1. Now we wish to At a ‘quadratic curve between Y and X as: ¥[n] = 01 + 0Xpn) + &sX%n] + Wn), where W[n] ~ (0.1) is the hite gaussian noise. Find out the best linear unbiased estimator for this problem. 6 Consider the observation model forthe unkuown parameter 7: X{n] =n" +1V[n].m = 0.1....,V~1, where Win] ~ N(O1) is the white gaussian noise, Derive the Newton-Raphson iterative algorithm to numerically compute the Maxianum Likelihood estimate of». ‘sl (& scanned with OKEN ScannerINDIAN INSTITUTE OF TECHNOLOGY, KHARAGPUR Mid-Autumn Semester 2022-23 Date of Examination : 2-|9\"2 session (FN/AN) FV _ Duration 2hrs Full Marks: 60 Subject No.: EC60501/£C41201 Subject: Diaital Image Processing /Image & Video Processing Depariment/Center/School: EZ ECE Specific charts, graph paper, log book etc., required Special Instructions (if any) : Answer all Questions, Answers to paris of a question must be in consecutive locations 4.(a) Ina Computer Vision Laboratory an experiment is set up to estimate the focal length (7.) of a camera. With great difficulty the students could come up with an experimental bench where a linear object of length 3 cm could be placed on the optical axis of the camera and perpendicular to it. It was observed that when the object was placed at certain distance from the camera it formed an image of height 1000 pixels. When the object was moved away from the camera by 36 cm the image height was reduced to 400 pixels. Assuming every pixel to be of size 51 x5yt, find out the focal length of the camera, (b) What is the maximum experimental error in your estimation? (c) Two identical cameras of focal length 60 mm are used in a stereo set up. The two cameras are only laterally displaced (along Y-axis) by 20 cm and imaging planes of bath the cameras are in the same plane. ‘There is no displacement along X or Z-axis. A point P in the 3D world is imaged at location P, (20, 50) in the left camera and at location P, (20, 70) in the right camera. Assuming every pixel to be of size 5415 1, find the location of P with respect to the left camera. 54347 = 15 2.(a) A 3x3 image P shown below is rotated around (3,3) by an angle 45° in the anti-clockwise direction to ‘form image Q (as shown in the shaded pixels). Compute the intensity values at pixels a, b, cand d in image Q using modified bilinear interpolation. Assume the top-left pixel at location (0,0). ° qa (b) State and explain Component Labelling Algorithm. PTO (& scanned with OKEN Scanner3. €@) Explain with diagram the operation of'an MPEG encoder. (b) The following matrix shows the quantized DCT coefficients of a 8x8 image macroblock. Assuming the DC coefficient ofthe previous macroblock to be 8, give the intermediate representation ofthe quantized DCT coefficients as per Baseline JPEG coding, 5-6 2 -1 0000 40 -1 000 30 0 2 0000 5 0 0 0 0 000 9 0 0 0 0000 9 0 0 0 0000 0 0 0 0 0000 0 0 0 0 0000 619-15 4. A transformation matrix A and an Image 1 is as given below. (2) Compute the Basis Images yo, 4y,, Ay and A,, for this transformation. (b) Show what will be the reconstructed ima; Avs Aas Avy and Ay, only. (©)Isita Unitary Transformation, why? ige 1, when the reconstruction is done using Basis Images os 05 05 05 1012 9 7 0.653 0.271 -0.271 -0.653 pe[5 19 33 12 05-05 -05 05 7 24 10 0.271 -0.653 0.653 -0.271 8 12 16 21 OHI=1S (& scanned with OKEN Scanner"INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR _ Mid-Autumn Semester Examination 2022-23 __ Date of Examined Session: (FN/AN) ___ Duration: 2.hrs_Full Mar Subject No.:_FoOv7it Subject :_ PATTERN Re GoGHTION AND MACIIE HTELUGENCE Departmenticentertseheot ESECE Specific charts, graph paper, log book etc., required _Not Required ______—— Special Instructions (if any): Answer AllQuestons vp = 0 to an arbitrary point x. = 0, assuming w #0. onto the above-mentioned 1. (a) Show that the distance from the hyperplane g(x) = Wx is |9(q)|/|wl] by minimizing {|x — xal|? subject to constraint 9(X) (b) From the above derivation, show that the projection of the points hyperplane is given by ec) Awl @ (a0+5) 2. (a) For n samples, consider a criterion function defined as: Jia) =F ey: by i and consider the following six training points: on: (8) 9) 5,9 wa 2-3) (A! 0,29" Calculate the Hessian matrix for this problem for the given sot of points, assuming a homo- ‘geneous representation of 4 85 Yi (b) Are these classes linearly separable? (c) Using the mentioned training set, classify a test point (3, yt k=3. Let using k-NN classifier considering (94+3+4) atablish the relation between MSE criterion and 6 ‘n-dimensional hypersphere with unit radius, the hypervolume in- Veyi(l) > Va(1); 0 € Ze. The symbols have their 3. @) Fisher’s Linear Discriminant. (b) Prove/Disprove that fo ‘ereases monotonically as n increases, 1.¢- ‘usual meanings. (©) An expression of within-group scatter S, of features y ‘ 3,=— — wy am z =o wy Show that S, can be written as 3, = (my ~ ma)? + sample mean and the scatter of class w,, respectively. (a) Prove/Disprove that. the eigenvectors corresponding to distinct eigenvalues of a real sym- metric matrix are orthogonal to each other. - € R can be given as 162+ Ls; where m, and s; are the (8+5+45+5) End of Paper (& scanned with OKEN ScannerINDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR Time: 2 hours, Full Marks: 60, Department of £ & ECE No. of students 60, Mid-Autumn Semester Examination, 2022 EC 60011, Data Structure and Object Representation Answer all questions: ig Oh: (33-3) swing by applying directly the definition 1. {a) Prove each of the follor (ii) nis O12) (nis (2) i) nifloga(n)} is fe") ion of n elements that are stored in the array A. This Permutation (Al J, n, 0). if Swap (A, i, i) 6) {b) The following C code gives the permutati is 2 recursive routine and is invoked by the call interchanges the i" andj" elements of A, calculate the running time of the routine. permutation (int A[ J, int m, int 2) ‘ int if = a) for (5-07 Jear j##) prints ("8d") RGGI)? printé ("\n"): else 7 deer 3) for (5 4 swap (A, i, DF Permutation (A, m, 342); swap (A, 3, 507 2. (a) Consider a prefixed expression. Using 2 stack to keep the operands, write a C function to ‘evaluate the expression. You may assume the stack operations push ) and pop{ ) 25 Wbrary functions. ” ment a Queve efficiently with the help of an array. Write a function ring this implementation. Show all the checks in the routine that is (44428) (b) Show haw you can impler to insert an element co required before any insertion. (a) Assume that there is a Circular Linked Listas follows: HOH Write a € function to find all nodes with element °X’in the list and replace them by element ‘Y a (& scanned with OKEN Scanner116 below. Convert it into a Right Pre-Threoded Binary Tree (b) Consider the binary tice in the ne its structure and write C functions for Setlett(p, ») by drawing appropriate dotted lines. De and Setright(p, x) for setting 1s left child and right child for any of ts nodes pointed to by 2. (2+6) 4. (a) Write a C type pseudo code that will sort elements kept in the nodes of @ linked list according, t0 Quick Sort algorithm. Give proper comments in the program to explain its various parts. (7) {6) Draw the Huffman Tree for the following numerals where their percentage probiilty of occurrences is given within the brackets: 0[25], [6], 2{22], 315], 415], 5(20}, 61281, 71231, 8{20], 92]. No codes necessary. Write a short C Code for deciphering the given sequence only. Assume ‘that p is the pointer to the Huffman Tree. (35) seneseeeeereeeeteenss BEST OF LUCK *Htb HO etetEereeneees (& scanned with OKEN ScannerINDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR mn AUTUMN 2022-2023, ination for the Ses Roll No. : DECESRI Name: Sor Koma Department ‘LECTRONICS & ELECTRICAL COMMUNICATION ENG, | Course Name = VISION AND INTELLIGENT SYSTEMS Hall Na ‘Suraj Kumar ‘Your Exam Time Schedule and Sitting Pattern Si | Sabuo. [Subject Name cNT@EPT| Date [Room | Time Span ‘Sitting 1 | EcéiROT [IMAGE AND WiDEO : i NAGE AND. ext | aezais [cies | onoormesoora [an 2 | Ece1203 [rouNDaTIONS OF LEARNING | ext. W (00 74-0500 Fy ® Founpa poaarai | NR oor 3 | Scer2nt |partenn RECOGNITION AND | ENT )) amzaiaa | Near | omoormosooma |e IEINTELLIGENCE™ 4 | Been [pata structures omsecr’ | parr | amatia3| Fan | ozoopnos.oo pat [REPRESENTATION eb s_|_rc«ts03[emmenven sysrexspesign | perr— | 2o7z.11as | roo | onaorncasoor [CNT Exam willie held Censally. DEFT: Faw bein ths Deprent Important Instructions and Guidelines During Examinations : 1, Entering to Examination all without valid Admit Card and Went Cards ot permite, 2, Mobil owes yo clearoni xis nchaig SMART WATCH ex inthe ited ot male, xp non pograiable clears, ' dresurciy BANNED ne Lxsmnion ia (Caryng any kindof cits, notebooks, books, oa ihr unauthorized document permit inthe Examination al Use 3. sappleenntary secu ns for ous wank Weltngon te qacton pera on hes card il esis ast mapa ease {ajone brings a ag, mut be hep ous he Examination Hal hs nak Sl ey panes lan pon Box iy sowed dent ae advised to occupy sat in he designs rom an mpi aig palin peaby 1S mines before ts schedule ea of he 4. amanon Itvny be ated atl Exon balou wil coeds misses tals Pheduedsring enc of coneoeion Bo om Sha 1:8 pm respectively for morning ard erent sao) accor wl be lowed § mms arte comencne of xsminaion Wik ‘Torough Freon and No ne wil be pemited to ener ti Exanvaon Hall minutes ater the commencement of Ux Gaminalion, Exchange of material such as calcula, pencil ea Se qiston pap, ot alloed 5. 6,1 n0 case, taking the answereript outside the Examination Halli permitted 77, Eneiog he Examination Ulin objectionable dress" mt pet. The ston fhe virile in this mater wl ef 8, Gathering information rom ay sure andr exchanngiformation with sor any such temps cite nse or ouside te Examination ‘al willbe west air mesa! le mean’ ser ese wile forwarded to Malpractice commie. This ma eat pushin ro eee ican year neers nnd dbarteat rom scilaahpy svar medals Signature of Student in fall Generation Date: Nov 10,2022 7:32 PM Generated By ERP. IT KOP (& scanned with OKEN Scanner=f, INDIAN INSTITUTE OF TECHNOLOGY KHARAGPUR Bees ___End-Autumn Semester Examination 2022-23 _ Date of Examinatio _____ Session: (FN/AN) Full Marks: 50 | Subject No: __EC61203_ Subject: Foundations of Learning Theory Department/Center/School: Department of E&ECE Specific charts, graph paper, log book etc,, required N/A Special Instructions (if any): Answer all questions. Make necessary assumptions if required. _ 1. Assume two independent continuous random variables U and V. Both of these random variables follow Uniform distributions with support [0,1], Now consider the following transformations: X = J=2nG) cos(2nv) ¥ = f=2in@W) sincanv) Prove that X and Y are Gaussian random variables. Are they independent and identically distributed? m 2. Consider a nonlinear signal model s{n; 0]. Linearize this signal around @ = Oy and derive the least square solution for computing the estimate 6. 8] 3. Show that the optimal Bayesian estimator is the median of the posterior density function if the Bayes Risk function is chosen as i R= Elo -6) ‘Where @ is the estimator of the parameter 0, am 4. Derive the Wiener-Hopf equation for estimating X(t-+ 1) given the known measurements (¥(e)) for 7 € Cat) (8) 5. Show that for Gaussian random variables X and Y, the conditional expectation E[X1Y] isa linear function of Y. (8) 6. For discrete time points T = {0,1,..~ 1,t,£ + 1), consider the following state space equations for a vector valued random process: X(t+1) (t+ 1;X(t) + UG) YO =MOXO x(0) =0 Where X(1) and U(z) are (n x 1) random vectors and ¥(r) isa (p 1) random vector for all € T, The state transition matrix ®(¢ + 15 ¢) isa (m x") stable fll rank time variant matrix. The observation matrix MCX) is 3 (p %n) ful rank time variant matrix where p