Akra–Bazzi method

In computer science, the Akra–Bazzi method, or Akra–Bazzi theorem, is used to analyze the asymptotic behavior of the mathematical recurrences that appear in the analysis of divide and conquer algorithms where the sub-problems have substantially different sizes. It is a generalization of the well-known master theorem, which assumes that the sub-problems have equal size.

The formula

The Akra–Bazzi method applies to recurrence formulas of the form

$T(x)=g(x) + \sum_{i=1}^k a_i T(b_i x + h_i(x))\qquad \text{for }x \geq x_0.$

The conditions for usage are:

• sufficient base cases are provided
• $a_i$ and $b_i$ are constants for all i
• $a_i > 0$ for all i
• $0 < b_i < 1$ for all i
• $\left|g(x)\right| \in O(x^c)$, where c is a constant and O notates Big O notation
• $\left| h_i(x) \right| \in O\left(\frac{x}{(\log x)^2}\right)$ for all i
• $x_0$ is a constant

The asymptotic behavior of T(x) is found by determining the value of p for which $\sum_{i=1}^k a_i b_i^p = 1$ and plugging that value into the equation

$T(x) \in \Theta \left( x^p\left( 1+\int_1^x \frac{g(u)}{u^{p+1}}du \right)\right)$

(see Θ). Intuitively, $h_i(x)$ represents a small perturbation in the index of T. By noting that $\lfloor b_i x \rfloor = b_i x + (\lfloor b_i x \rfloor - b_i x)$ and that $\lfloor b_i x \rfloor - b_i x$ is always between 0 and 1, $h_i(x)$ can be used to ignore the floor function in the index. Similarly, one can also ignore the ceiling function. For example, $T(n) = n + T \left(\frac{1}{2} n \right)$ and $T(n) = n + T \left(\left\lfloor \frac{1}{2} n \right\rfloor \right)$ will, as per the Akra–Bazzi theorem, have the same asymptotic behavior.

An example

Suppose $T(n)$ is defined as 1 for integers $0 \leq n \leq 3$ and $n^2 + \frac{7}{4} T \left( \left\lfloor \frac{1}{2} n \right\rfloor \right) + T \left( \left\lceil \frac{3}{4} n \right\rceil \right)$ for integers $n > 3$. In applying the Akra–Bazzi method, the first step is to find the value of p for which $\frac{7}{4} \left(\frac{1}{2}\right)^p + \left(\frac{3}{4} \right)^p = 1$. In this example, p = 2. Then, using the formula, the asymptotic behavior can be determined as follows:

$\begin{align} T(x) & \in \Theta \left( x^p\left( 1+\int_1^x \frac{g(u)}{u^{p+1}}\,du \right)\right) \\ & = \Theta \left( x^2 \left( 1+\int_1^x \frac{u^2}{u^3}\,du \right)\right) \\ & = \Theta(x^2(1 + \ln x)) \\ & = \Theta(x^2 \log x). \end{align}$

Significance

The Akra–Bazzi method is more useful than most other techniques for determining asymptotic behavior because it covers such a wide variety of cases. Its primary application is the approximation of the runtime of many divide-and-conquer algorithms. For example, in the merge sort, the number of comparisons required in the worst case, which is roughly proportional to its runtime, is given recursively as $T(1) = 0$ and

$T(n) = T\left(\left\lfloor \frac{1}{2} n \right\rfloor \right) + T\left(\left\lceil \frac{1}{2} n \right\rceil \right) + n - 1$

for integers $n > 0$, and can thus be computed using the Akra–Bazzi method to be $\Theta(n \log n)$.

References

• Mohamad Akra, Louay Bazzi: On the solution of linear recurrence equations. Computational Optimization and Applications 10(2):195–210, 1998.
• Tom Leighton: Notes on Better Master Theorems for Divide-and-Conquer Recurrences, Manuscript. Massachusetts Institute of Technology, 1996, 9 pages.
• Proof and application on few examples