6.2-final-2010-strang¶
Final Exam, 2010 (strang) · Section 6.2: Diagonalizing a Matrix Source: finalexam_2.pdf, pp. 29-30
Problem¶
4 (11 points)
(a) Suppose \(x_k\) is the fraction of MIT students who prefer calculus to linear algebra at year \(k\). The remaining fraction \(y_k = 1 - x_k\) prefers linear algebra. At year \(k + 1\), 1/5 of those who prefer calculus change their mind (possibly after taking 18.03). Also at year \(k + 1\), 1/10 of those who prefer linear algebra change their mind (possibly because of this exam). Create the matrix \(A\) to give
and find the limit of \(A^k \begin{bmatrix} 1 \\ 0 \end{bmatrix}\) as \(k \to \infty\).
(b) Solve these differential equations, starting from \(x(0) = 1, y(0) = 0\):
(c) For what initial conditions \(\begin{bmatrix} x(0) \\ y(0) \end{bmatrix}\) does the solution \(\begin{bmatrix} x(t) \\ y(t) \end{bmatrix}\) to this differential equation lie on a single straight line in \(\mathbb{R}^2\) for all \(t\)?
Solution
(a) The Markov matrix is \[ A = \begin{bmatrix} .8 & .1 \\ .2 & .9 \end{bmatrix}. \] The eigenvector with \(\lambda = 1\) is \((1/3, 2/3)\): this is the steady state, and the limit from any start (in particular from \((1, 0)\)) is \[ \lim_{k \to \infty} A^k \begin{bmatrix} 1 \\ 0 \end{bmatrix} = \begin{bmatrix} 1/3 \\ 2/3 \end{bmatrix}. \] So eventually 2/3 of all students prefer linear algebra! I agree. (b) The matrix \[ A = \begin{bmatrix} 3 & -4 \\ 2 & -3 \end{bmatrix} \] has eigenvalues \(\lambda_1 = 1\) and \(\lambda_2 = -1\) with eigenvectors \(x_1 = (2, 1)\) and \(x_2 = (1, 1)\). The initial vector \((x(0), y(0)) = (1, 0)\) is \(x_1 - x_2\). So the solution is \[ (x(t), y(t)) = e^t (2, 1) + e^{-t} (1, 1). \] (c) If the initial conditions are a multiple of either eigenvector \((2, 1)\) or \((1, 1)\), the solution is at all times a multiple of that eigenvector, so it lies on a single straight line in \(\mathbb{R}^2\) for all \(t\).Keywords: Markov matrix, steady state, eigenvectors, differential equations