This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Applied linear algebra and numerical analysis Session 18 Prof. Ulrich Hetmaniuk Department of Applied Mathematics February 17, 2010 Operation counts • The work required to solve a problem on a computer is often measured in terms of the number of fl oating point op erations or flops needed to do the calculation. A floating point operation is multiplication, division, addition or subtraction on floating point numbers (real numbers represented on the computer). Operation counts Example Computing the inner product of two vectors in R n , x T y = x 1 y 1 + x 2 y 2 + ··· + x n y n , requires n multiplications and n 1 additions, so 2 n 1 flops. • Computing an inner product will take about 2 times more operations for a vector with 1000 rows than for a vector with 500 rows. • It should take about 2 times more seconds. Operation counts Example Computing y = Ax where A ∈ R m × n and x ∈ R n . The ith element of y is the inner product of the ith row of A with x and requires 2 n 1 flops. There are m rows. So we need to compute y i for i = 1 , 2 ,..., m . The total work is 2 mn m flops. When the matrix A is square ( m = n ) , the count becomes 2 n 2 n . • Computing a matrixvector product will take about 4 times more operations (or longer) for a 1000 × 1000 matrix than for a 500 × 500 matrix. Operation counts Example Consider A ∈ R m × r and B ∈ R r × n . Computing the matrixmatrix product C = AB will require the computations of mn entries c ij . Each entry c ij is the inner product of two vectors with r components, which require 2 r 1 flops. So the total work becomes 2 mnr mn flops. • Computing a matrixmatrix product will take about 8 times more longer for a 1000 × 1000 matrix than for a 500 × 500 matrix. “Big oh” notation Definition The function W ( n ) is “Big oh” of n k when the ratio W ( n ) / n k remains bounded as n→ + ∞ : W ( n ) = O ( n k ) ⇔  W ( n )  ≤ Cn k for large values of n • If the work required by some algorithm for a system of size...
View
Full
Document
This note was uploaded on 03/31/2010 for the course AMATH 352 taught by Professor Leveque during the Winter '07 term at University of Washington.
 Winter '07
 Leveque
 Numerical Analysis

Click to edit the document details