Suppose I have a case like
T(n)=2T(n/4)+log(n). a=2, b=4, f(n)=log(n)
That should be case 1 because n^(1/2)>log(n)
. There is also a lambda in case 1. f(n)=O(n^((1/2)-lambda)
. Is this correct? And how can I find this lambda?
Suppose I have a case like
T(n)=2T(n/4)+log(n). a=2, b=4, f(n)=log(n)
That should be case 1 because n^(1/2)>log(n)
. There is also a lambda in case 1. f(n)=O(n^((1/2)-lambda)
. Is this correct? And how can I find this lambda?
The constant lambda is important: its purpose is to avoid considering the bizarre cases that lie between Case 1 and Case 2. Since big-O is an upper bound only and not a lower bound, smaller choices of lambda are "better" in the sense that they cover more functions. Since lambda must be positive, however, there is no "best" choice of lambda. Lambda = 10^-3 should get you through enough examples to see why most treatments of the Master Theorem do not make a production out of choosing lambda.