TOPICS
Search

Akra-Bazzi Method


The Akra-Bazzi method determines the asymptotic growth of running-time recurrence equations arising from divide-and-conquer algorithms. In one common form, the recurrence is

 T(x)=sum_(i=1)^ka_iT[b_ix+h_i(x)]+g(x),
(1)

where a_i>0, 0<b_i<1, the perturbations satisfy |h_i(x)|=O(x/ln^2x), and the function g(x) obeys the required regularity and polynomial-growth conditions. If p is the unique real number satisfying

 sum_(i=1)^ka_ib_i^p=1,
(2)

then

 T(x)=Theta(x^p[1+int_1^x(g(u))/(u^(p+1))du]).
(3)

The method extends the familiar equal-subproblem result for a divide-and-conquer algorithm by allowing unequal subproblem sizes and controlled perturbations of their arguments (Akra and Bazzi 1998).


See also

Asymptotic Notation, Divide-and-Conquer Algorithm, Recurrence Equation

Explore with Wolfram|Alpha

References

Akra, M. and Bazzi, L. "On the Solution of Linear Recurrence Equations." Comput. Optim. Appl. 10, 195-210, 1998. https://doi.org/10.1023/A:1018373005182.

Cite this as:

Weisstein, Eric W. "Akra-Bazzi Method." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/Akra-BazziMethod.html

Subject classifications