How to solve T(n)=4T(sqrt(n/2))+n^(3/2)

The time function is:

enter image description here

We can make this substitution:

enter image description here

And thus:

enter image description here

Now the i + 1th expansion of the time function gives a term:

enter image description here

And the termination index of i, assuming T(0) = 0:

enter image description here

And so the time complexity is given by:

enter image description here

Unfortunately, this is non-analytical (no elementary function representation).

Instead, however, we can make an extremely crude approximation which will almost certainly flabbergast every math-mo out there. Taking the terms at either end of the sum:

enter image description here

Assuming m is large, the first term is much much larger than the last. So we can make a lower bound on the series:

enter image description here


EDIT: Apologies, the sum starts from i = 0, including the last 2^(1.5*m) term. BUT the time complexity bound is still the same. Also I should have used big-Omega notation not big-O…

Leave a Comment