Technische Universität München Zentrum Mathematik

Similar documents
Technische Universität München Zentrum Mathematik

MATH10212 Linear Algebra B Homework 7

3 Matrix Algebra. 3.1 Operations on matrices

Lesson 3. Inverse of Matrices by Determinants and Gauss-Jordan Method

Conics and their duals

Eigenvalues and Eigenvectors

MATH 2050 Assignment 8 Fall [10] 1. Find the determinant by reducing to triangular form for the following matrices.

Lecture 2 Systems of Linear Equations and Matrices, Continued

MATRIX ALGEBRA AND SYSTEMS OF EQUATIONS. + + x 1 x 2. x n 8 (4) 3 4 2

Linear Algebra. Min Yan

Choose three of: Choose three of: Choose three of:

Linear Algebra: Lecture Notes. Dr Rachel Quinlan School of Mathematics, Statistics and Applied Mathematics NUI Galway

Final Review Sheet. B = (1, 1 + 3x, 1 + x 2 ) then 2 + 3x + 6x 2

Honors Advanced Mathematics Determinants page 1

30.3. LU Decomposition. Introduction. Prerequisites. Learning Outcomes

Methods for Solving Linear Systems Part 2

SPRING OF 2008 D. DETERMINANTS

1111: Linear Algebra I

MATRIX DETERMINANTS. 1 Reminder Definition and components of a matrix

Chapter 5: Matrices. Daniel Chan. Semester UNSW. Daniel Chan (UNSW) Chapter 5: Matrices Semester / 33

Lecture 8: Determinants I

1 Last time: determinants

Section 1.5. Solution Sets of Linear Systems

Extra Problems for Math 2050 Linear Algebra I

Topic 15 Notes Jeremy Orloff

Hermite normal form: Computation and applications

MIDTERM 1 - SOLUTIONS

Queens College, CUNY, Department of Computer Science Numerical Methods CSCI 361 / 761 Spring 2018 Instructor: Dr. Sateesh Mane.

Chapter 4. Solving Systems of Equations. Chapter 4

7.6 The Inverse of a Square Matrix

SAMPLE OF THE STUDY MATERIAL PART OF CHAPTER 1 Introduction to Linear Algebra

Second Midterm Exam April 14, 2011 Answers., and

Elementary Linear Algebra

MTH Linear Algebra. Study Guide. Dr. Tony Yee Department of Mathematics and Information Technology The Hong Kong Institute of Education

1300 Linear Algebra and Vector Geometry

Linear Algebra 1 Exam 1 Solutions 6/12/3

1111: Linear Algebra I

Technische Universität München, Zentrum Mathematik Lehrstuhl für Angewandte Geometrie und Diskrete Mathematik. Combinatorial Optimization (MA 4502)

October 25, 2013 INNER PRODUCT SPACES

1 Differentiable manifolds and smooth maps. (Solutions)

Mobile Robotics 1. A Compact Course on Linear Algebra. Giorgio Grisetti

APPLICATIONS The eigenvalues are λ = 5, 5. An orthonormal basis of eigenvectors consists of

2018 Fall 2210Q Section 013 Midterm Exam I Solution

Math 250B Midterm I Information Fall 2018

MATH10212 Linear Algebra B Homework 6. Be prepared to answer the following oral questions if asked in the supervision class:

MTH 464: Computational Linear Algebra

Digital Workbook for GRA 6035 Mathematics

The complex projective line

1 Differentiable manifolds and smooth maps. (Solutions)

MATRICES. knowledge on matrices Knowledge on matrix operations. Matrix as a tool of solving linear equations with two or three unknowns.

4.2. ORTHOGONALITY 161

Maths for Signals and Systems Linear Algebra in Engineering

Math Computation Test 1 September 26 th, 2016 Debate: Computation vs. Theory Whatever wins, it ll be Huuuge!

Solution to Homework 1

Matrix Factorization and Analysis

Lecture Summaries for Linear Algebra M51A

Math 54 HW 4 solutions

Lecture 9: Elementary Matrices

Introduction to Systems of Equations

1 Determinants. 1.1 Determinant

1 Last time: least-squares problems

MODEL ANSWERS TO HWK #3

22.3. Repeated Eigenvalues and Symmetric Matrices. Introduction. Prerequisites. Learning Outcomes

Matrices. In this chapter: matrices, determinants. inverse matrix

ECON 186 Class Notes: Linear Algebra

Linear Algebra: Matrix Eigenvalue Problems

5.6. PSEUDOINVERSES 101. A H w.

SAMPLE OF THE STUDY MATERIAL PART OF CHAPTER 1 Introduction to Linear Algebra

ANALYTICAL MATHEMATICS FOR APPLICATIONS 2018 LECTURE NOTES 3

LS.1 Review of Linear Algebra

11 a 12 a 13 a 21 a 22 a b 12 b 13 b 21 b 22 b b 11 a 12 + b 12 a 13 + b 13 a 21 + b 21 a 22 + b 22 a 23 + b 23

Matrices. Chapter Definitions and Notations

MATH 213 Linear Algebra and ODEs Spring 2015 Study Sheet for Midterm Exam. Topics

Math 462: Homework 2 Solutions

MATH 110: LINEAR ALGEBRA PRACTICE MIDTERM #2

Algebraic. techniques1

Modeling and Simulation with ODE for MSE

Getting Started with Communications Engineering

MTH 35, SPRING 2017 NIKOS APOSTOLAKIS

1 procedure for determining the inverse matrix

Law of Trichotomy and Boundary Equations

Phys 201. Matrices and Determinants

MATH 2210Q MIDTERM EXAM I PRACTICE PROBLEMS

ANSWERS (5 points) Let A be a 2 2 matrix such that A =. Compute A. 2

Further Mathematical Methods (Linear Algebra) 2002

FACULTY OF MATHEMATICAL STUDIES MATHEMATICS FOR PART I ENGINEERING. Lectures

Introduction to Geometry

Linear Equation: a 1 x 1 + a 2 x a n x n = b. x 1, x 2,..., x n : variables or unknowns

Multiplying matrices by diagonal matrices is faster than usual matrix multiplication.

MH1200 Final 2014/2015

Gaussian Elimination and Back Substitution

TBP MATH33A Review Sheet. November 24, 2018

Basic Surveying Week 3, Lesson 2 Semester 2017/18/2 Vectors, equation of line, circle, ellipse

EXAM 2 REVIEW DAVID SEAL

Overview. Motivation for the inner product. Question. Definition

Numerical Methods Lecture 2 Simultaneous Equations

LINEAR ALGEBRA REVIEW

Math/CS 466/666: Homework Solutions for Chapter 3

(Refer Slide Time: )

Lines and points. Lines and points

Transcription:

Technische Universität München Zentrum Mathematik Prof. Dr. Dr. Jürgen Richter-Gebert, Bernhard Werner Projective Geometry SS 8 https://www-m.ma.tum.de/bin/view/lehre/ss8/pgss8/webhome Solutions for Worksheet 9 (8-6-8) Question. Intersecting conics Classwork Let the following two matrices describe two conics: 3 A = 3 B = 8 4 The goal of this task is to determine all points of intersection between these two conics. a) Determine a linear combination C = A + λb which represents a degenerate conic, and compute its matrix. b) Transform this matrix C into an equivalent matrix C of rank. c) Read the constituent lines g and h from this matrix. d) Determine a matrix M g in such a way that for every vector v R 3 the equation M g v = g v holds. Also compute M h in the same way. If you don t remember the structure of such a matrix by heart, try to derive it instead of looking it up in your lecture notes. e) Find the matrix D of a degenerate dual conic that has all those lines as tangents which pass through either of the points of intersection between g and A. f) Find a matrix D equivalent to D but of rank. Make use of the fact that you know the homogeneous coordinates of g. Read the coordinates of two points of intersection from this matrix. g) In the same fashion, compute the points of intersection between line h and conic B. h) Write down a list of all the points during this computation where you either made an arbitrary choice, or where the problem statement made an arbitrary choice, even though a computation based on a different choice would have lead to the same final results. i) Identify the other degenerate conics which could result from subtask a), and split at least one of them into lines k and l. j) Compute all those points of intersection between the lines g, h, k, l which are also points of intersection between the conics. Use this to verify your previous results.

Solution: a) A conic is degenerated if the determinant of its symmetric matrix is zero. This leads to: 3 = det(a + λb) = det 3 + λ 8 4 3 λ λ = 3 λ λ λ λ 4(λ ) 3 λ λ = (λ ) 3 λ λ 4 At this point we already see that λ = has to be a root of the determinant, which is enough for this part. C = A + B = b) Because of the very easy structure of C one can spot a suitable rank--matrix basically immediately: C = C + = c) A non-zero row or column of the matrix is a representative of one of the factor lines. C = = (,, ) g = h = d) If you do not know the structure of such a matrix M g by heart, reconstruct it from an example: v v g v = v = v 3 = v = M g v M g = v 3 v v 3 v v 3 v h v = v = = v = M h v M h = v 3 v v 3 e) A line m is tangential to D iff the intersection of m and g lies on A. (g m) T A (g m) = m T M T g A M g m We can deduce the matrix of the conic from this: D = M T g A M g = 3 3 8 = 8 = 8 3

f) The matrix D must result as a linear combination of D and M g, as g is the connecting line of the two points described by the dual conic to D. D = D + λm g = 8 + λ In this easy case, one can see without computing the determinant that λ = ± leads to a rank--matrix. D = D + M g = 8 + = 8 4 D = 4 (,, ) This leads to the two intersections of g and A: p = p = g) Analogously we get ther intersections of h and B: E = M T h B M h = 4 4 4 = = 4 E = E + M h = = (,, ) p 3 = h) At the following points we made an arbitrary choice: p 4 = In the linear combination in part a): Which conic has the coefficient. Another possibility would have been C = B + λa. In part a): Which of three roots of the determinant to use. These roots correspond to the three degenerated conics through the four points. In part b) and f), we could have used the transposed matrix or could have done without the scaling. The assignment of the letters g an h to rows and columns is arbitrary. The other assignment leads to swapping the lines just as transposing the matrix does. As we explicitly use g in the succeeding computations, this choice influences the next steps, but not the entirety of the intersection points. It does not matter which line and which conic to intersect. One can intersect both with A, both with B or g with B ans h with A. 3

i) To determine the other degenerated conics, we have to expand the determinant from part a). 3 λ λ det(a + λb) = (λ ) 3 λ λ 4 ( = (λ ) (λ 3) 3 λ λ ) 4 3 λ λ λ = (λ ) (λ 3) ((3 λ) 4 ( λ) ( ) (λ ) ) = (λ ) (λ 3) ( 4λ + λ + λ ) = (λ ) (λ 3) (8 λ) = 4 (λ ) (λ 3) (λ 4) With this, we know the other roots. And the respective conics unfold as linear combinations: A + 3B = = (,, ) 4 4 A + 4B = 4 = (,, ) 8 4 8 k = l = j) We can the four intersection points also by the following computation: g = h = k = l = g k = = p g l = = p h k = = p 4 h l = = p 3 This ansatz yields the same four intersection points (up to scalars) as above. To close this exercise, a picture to visualise the conics and their finite and infinite intersections: A B 4

Question. Decomposing a conic Decompose the following conic into its constituent lines: 4 A = 4 7 7 8 Solution: First, we determine the dual conic B = A = 4 7 7 8 7 8 4 7 7 8 4 8 4 7 4 7 4 7 4 4 8 4 7 9 6 3 = 4 36 8 6 4 7 8 9 3 From this matrix we can infer (by considering any non-zero row or column) that the intersection of the two lines is p = (3,, ) T. So it would have been enough to compute a single row or column. When looking a the structure of the adjugate matrix, we see that a single column is computed via -determinants with alternating sign. We recall this structure from the cross product. So, computing directly from A: 4 7 3 4 = 8 = p 7 9 The anti-symmetric matrix we have to add to A to get rank is therefore a multiple of C = M p = 3 3 Now we have to find a suitable multiple. When A + αc has rank, every -determinant will be zero. Using the bottom right sub-matrix, we get 4 7 + 3α 7 3α 8 = 4 7 4 7 7 8 8 7 8 + (3α) = α = ± = ± = ±3 3 3 Of course, we could have used any of the other two -determinants with the top left one we would even have avoided division. What becomes apparent with the above solution is, however, that the respective coordinate of p appears in the denominator of the solution for α. The corresponding determinant appears in the numerator. This determinant we could also read from the adjugate. With the proportion factor α = 3 we found, we can compute the rank--matrix: D = A + 3C = 4 + 3 6 4 8 4 3 4 7 + 9 = 4 = (,, ) + 6 7 9 8 8 6 8 4 So, the two lines have the homogeneous coordinates. g = h = 4

Question 3. From where was a picture taken Given the following points in RP Homework A = B = 3 C = D = E = you are asked to compute the point F RP (distinct from A through E) which satisfies (A, B; C, D) F = (B, C; D, E) F = 3 a) Each of these cross ratios defines a conic. The value has special significance for cross ratios. What is special about the corresponding conic, and how can you exploit this fact in subsequent tasks? b) Find matrices which represent the two conics for the given cross ratios. c) Find the point F which satisfies the above condition. Solution: a) A cross-ratio of, or in RP means that at least two of the points are the same. Concretely, for (A, B; C, D) = we have A B or C D. This is obvious from the definition of the cross-ratio. Transferring this to the plane viewed from a fifth point F, this means that the respective connecting lines collapse. I.e. either ABF are collinear or CDF. So the conic degenerates into two lines. b) The conic corresponding to the first equation can be directly determined from the two lines. (A B) (C D) T = 3 A suitable symmetric representation would be 6 6 Q + Q T = 6 T 3 3 = (,, ) = = Q 6 6 For the conic with cross-ratio 3 it is best to orientate oneself towards exercise on sheet 8. First, we have to determine the conics with cross-ratio, and. (B D) (C E) T = (,, ) = R 4 (B C) (D E) T = (,, ) = = R 4 4 (B E) (C D) T = (,, ) = = R 7 7 6

These conics should be symmetrised to be able to handle the next step. S = R + R T = 8 S = R + R T = 8 7 S = R + R T = 7 The next step is to adjust the scale defined by these three conics we need representatives for S and S such that their sum is (a multiple of) S. Because of the zeros we can read the respective factors directly from the top left and bottom right corner: S + S = S ( So we found the basis matrices: S takes on the role of and S ) the role of ( 3 this scale we get by a point. ) ( ). The cross-ratio of 3 for 8 7 4 6 6 3 S + S = 3 + = 6 = K 7 6 c) In theory, we would now have to intersect the conics Q and K. We know, however, that Q is degenerated and we even know its components. Furthermore, we know that Q and K have the points B, C and D in common. It thus suffices to intersect the line AB with the conic K. 3 A B = 6 6 M A B = 6 3 3 6 4 6 6 6 L = M T A B K M A B = 6 3 6 6 3 3 6 3 6 = 66 6 6 3 = 8 3 6 6 3 3 This matrix describes the two intersection points of the line AB with the conic K. To decompose it into these points we have to add an anti-symmetric matrix to get a rank--matric. To be precise, we have to add a suitable multiple of M A B. We can get the proportion factor for example from the minor obtained by deleting the middle row and middle column. This has the advantage that the middle coordinate of A B is and we therefore do not have to divide by it. α = = = 4 + = 3 L + α M A B = 3 8 = 3 (, 3, ) 3 8 3 So, the conic K intersects the line AB in the known point B and the desired point F. It is therefore determined as F = 3 3 7

The computation to get this final result was quite long and some intermediate result are still uncomfortably large. Moreover, we had to take care of many signs. This makes computational errors very likely. So it makes sense to verify the result at this point. Computing the cross-ratios given in the instructions with the computed F does result in the given values. The result is correct. Question 4. Degrees of freedom of conics a) How many real degrees of freedom does one have in choosing an arbitrary conic in RP? b) Suppose you have chosen such a conic. Now we are looking for projective transformations which map the unit circle to this conic. How many real degrees of freedom are there to choose such a transformation? In subsequent subtasks this figure shall be denoted as k. c) Suppose you are given k pairwise distinct points P i on the unit circle, and also k pairwise distinct points P i on a conic represented by some matrix B. Describe how one could obtain the matrix of the projective transformation which maps the unit circle onto B. d) Execute this computation using specific numbers. You are free to choose coordinates in such a way that the computations become as simple as possible. The conic B shall not be degenerate, though, and not a circle either. Furthermore, you should perform all steps as you described them in c), even though your specific choice of coordinates could make some steps more simple. You may of course use these easier solutions to verify your results. Solution: a) A conic is described by an equation with 6 coefficients. This corresponds to the 6 entries of a triangular matrix, from which the other other entries of a symmetric matrix yield. A multiple of such an equation or matrix gives, however, the same conic. This leaves degrees of freedom. An alternative approach is the following: A conic can be described by points in the plane. Each point gives real degrees of freedom. But each of these points can be shifted along the conic without changing the conic. This eliminates degree of freedom. This leaves degrees, too. b) A conic is parametrised by RP. Hence, three points on the conic fix a projective scale. All other points are uniquely determined by their cross-ratio w.r.t. these three points. Each of these three base points can be chosen on the conic, giving degree of freedom each. I.e. 3 in total. c) One possibility is to construct the point P 4 which is in harmonic position w.r.t. P through P 3 i.e. satisfying (P, P ; P 3, P 4 ) =. It can be determined, for example, by Hesse s transfer principle. Analogously, we can determine a point P 4 as its image. Then we have four points and their images which determine a unique projective transformation, as we know. d) There are many reasonable choices for suitable numbers. Furthermore, there is a certain trade-of between easy computations and a high learning effect. The more complicated the coordinates are the better one has to understand the general approach. This solution treats a rather easy case: The transformation which is the final result will be given in advance. For example, one could choose the transformation that scales every x-coordinate by. The image of the unit circle would then be an ellipse with its center of symmetry at the origin. As defining points one could use points on the coordinate axes. The corresponding details would like this: A = B = P = P = P = P = P 3 = P 3 = 8

With this we can compute the points P 4 and P 4. To do so, we take the tangents at the conic in P and P, connect their intersection with P 3 and intersect this line with the conic. Starting with the intersection of the tangents: Q (A P ) (A P ) = = = Q 8 Q (B P ) (B P ) = = = Q And then... P 4 = Q T AQ P 3 P3 T AQ Q = P 3 Q = P 4 = Q T BQ T P 3 P 3 T BQ Q = P 3 Q = Now we need transformations mapping the standard basis of RP onto the corresponding point quadruple. λ λ µ = µ = τ τ λ λ µ = µ = τ τ M = M = Of these maps, M will gets executed in the opposite direction and afterwards M in forward direction. The compound map is therefore 4 M = M M = = That is exactly the transformation we chose in the beginning. This means the computation is correct. It can be applied to other points and conics by using other numbers. The method is sufficiently general. 9