Scribd Logo
Upload

Welcome to Scribd!

  • 104 views

    Ict 4052 Nnfl-Mkp-Part B

    Uploaded by

    Shan
    1. This document outlines exam questions for a course on Neural Networks and Fuzzy Logic. 2. Question 1 asks students to write the Lagrangian and dual optimization problem for 2 norm soft margin SVM (part A), find the optimum value of w for a given cost function E(w) (part B), and discuss differences between Bayes classifier and perceptron (part C). 3. Question 2 asks students to show the first iteration of backpropagation for a given dataset using a 3x3x2 neural network (part A), perform max-min and max-product composition on given relations R and S (part B), and draw the structure of an RBF network and describe its components (part C).

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Ict 4052 Nnfl-Mkp-Part B

    Uploaded by

    Shan
    104 views2 pages
    1. This document outlines exam questions for a course on Neural Networks and Fuzzy Logic. 2. Question 1 asks students to write the Lagrangian and dual optimization problem for 2 norm soft margin SVM (part A), find the optimum value of w for a given cost function E(w) (part B), and discuss differences between Bayes classifier and perceptron (part C). 3. Question 2 asks students to show the first iteration of backpropagation for a given dataset using a 3x3x2 neural network (part A), perform max-min and max-product composition on given relations R and S (part B), and draw the structure of an RBF network and describe its components (part C).

    Original Title

    ICT 4052 NNFL-MKP-PART B

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    1. This document outlines exam questions for a course on Neural Networks and Fuzzy Logic. 2. Question 1 asks students to write the Lagrangian and dual optimization problem for 2 norm soft margin SVM (part A), find the optimum value of w for a given cost function E(w) (part B), and discuss differences between Bayes classifier and perceptron (part C). 3. Question 2 asks students to show the first iteration of backpropagation for a given dataset using a 3x3x2 neural network (part A), perform max-min and max-product composition on given relations R and S (part B), and draw the structure of an RBF network and describe its components (part C).

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    104 views2 pages

    Ict 4052 Nnfl-Mkp-Part B

    Uploaded by

    Shan
    1. This document outlines exam questions for a course on Neural Networks and Fuzzy Logic. 2. Question 1 asks students to write the Lagrangian and dual optimization problem for 2 norm soft margin SVM (part A), find the optimum value of w for a given cost function E(w) (part B), and discuss differences between Bayes classifier and perceptron (part C). 3. Question 2 asks students to show the first iteration of backpropagation for a given dataset using a 3x3x2 neural network (part A), perform max-min and max-product composition on given relations R and S (part B), and draw the structure of an RBF network and describe its components (part C).

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 2

    Reg. No.

    SEVENTH SEMESTER B.TECH (INFORMATION TECHNOLOGY / COMPUTER AND


    COMMUNICATION ENGINEERING) DEGREE MAKEUP EXAMINATION-FEB 2022
    SUBJECT:PROGRAM ELECTIVE-V NEURAL NETWORKS AND FUZZY LOGIC (ICT 4052)
    (REVISED CREDIT SYSTEM)
    TIME: 1.15 HOURS 17/02/2022 MAX. MARKS: 20

    Instructions to candidates
    ˆ Answer ALL questions.
    ˆ Missing data, if any may be suitably assumed.

    PART-B

    1A. For `2 norm soft margin SVM, the algorithm is given by the following optimization prob-
    lem: m
    1 2 CX 2
    min ||w|| + ξ
    w,b,ξ 2 2 i=1 i
    s.t. y (i) (wT x(i) + b) ≥ 1 − ξi i = 1, . . . , m

    i) Write the Lagrangian for the given optimization problem.


    ii) Derive the dual optimization problem.
    [5]

    1B. Consider the cost function


    1 1
    E(w) = σ 2 − rTxd w + wT Rx w
    2 2
    where σ 2 is some constant and
       
    0.8182 1 0.8182
    rxd = , Rx =
    0.354 0.8182 1

    Find the optimum value of w∗ for which E(w) reaches its minimum value. [3]

    1C. The opeartion of the Bayes classifier for the Gaussian environment is analogous to that
    of the perceptron in that they are both linear classifiers. However, there are some subtle
    and important difference between them. Briefly discuss those differences.
    [2]
    2A. For the data shown in Table Q.2A, show the first iteration of back propagation algorithm
    in trying to compute the membership values for the input variables x1 , x2 and x3 in the
    output region R1 and R2 . Use a 3 × 3 × 2 neural network architecture. Assume a random
    set of weights for your neural network.

    Table: Q.2A

    x1 x2 x3 R1 R2
    1.5 1.0 0.8 1.0 0.0
    [5]

    ICT 4052 Page 1 of 2


    2B. In photography, it is important to relate reagent thickness to color balance on the film
    medium. Let Y be a universe of color balance, Y = [0, 1, 2, 3, 4], where 0=yellow, 4=blue,
    and 2=neutral. Let X be a universe of the reagent thickness, X = [0, 1, 2, 3, 4], where
    0=thin, 4=thick, and 2=semi-thick. Now, suppose that a relation is obtained from a
    Cartesian product as follows:

    0 1 2 3 4
    0 1
     0.8 0.6 0.2 0 
    1 0.8 1 0.8 0.6 0 
    R=X ×Y = 2 
    0.6 0.8 1 0.8 0.6

    3 0.2 0.6 0.8 1 0.8
     
    4 0 0.2 0.6 0.8 1

    It is required to relate color balance on the film medium to the perceived quality of
    the picture. For this we need an additional universe of perceived picture quality, Z =
    [0, 1, 2, 3, 4], where 0=bad, 4=excellent, and 2=fair. Suppose a relation is obtained from
    a Cartesian product

    0 1 2 3 4
    0  1 0.6 0.4 0.2 0 
    1 0.6 1 0.6 0.4 0 
    S =Y ×Z = 2 0.4 0.6 1

    0.6 0.4

    3 0.2 0.4 0.6 1 0.6
     
    4 0 0.2 0.4 0.6 1

    i) Find T = R ◦ S using max-min composition


    ii) Find T = R ◦ S using max-product composition.
    [3]
    2C. Draw the structure of an RBF network from interpolation theory perspective and briefly
    write about the various components of this network. [2]

    ICT 4052 Page 2 of 2

    You might also like