1

I am confused with respect to this equality. For $c>=1$, $(n-cn)$ should be less and less. For example if $c = 2, then (n-cn)$ should be $-n$ and overall equality changes to $T(n)<= nlogn - n$. And thus $cn\log n-cn+n$ should decrease for larger value of $c$ and the <= should not hold, as $cn\log n-cn+n$ decreases as $c$ increase. I know for a fact that this is correct, so where am I wrong?

2 Answers2

1

Since $c\ge 1$ you have $1-c \le 0$ and therefore $$cn\lg n - cn+n = cn \lg n + n(1-c) \le cn \lg n$$

gammatester
  • 18,827
0

By the way, your question is rather confusingly written. For $c>1$

$T(n) \le cn \lg n -cn+n \le cn \lg n$ holds for all $n$.

Meanwhile $cn \lg n$ increases faster than $-cn$ decreases (for large $n$). And as $n$ tends to infinity, $cn \lg n$ increases (with $c$) much faster than $-cn$ decreases.

The $cn \lg n$ term dominates the $cn$, $n$ terms.

Mike
  • 20,434
  • How do i prove it. I am using induction and $T(n) <= cn lgn - cn + n$ is given to me. Logically speaking $cn lgn - cn + n$ wil decrease for all c>=1. So, We cant say that $T(n) <= cn lgn - cn + n$ holds for all c >=1. I know it does but how to prove it mathematically. – trickster Jun 03 '18 at 21:57
  • No it will not decrease, $cn \lg n$ increases w $c$, faster than $-cn$ decreases. If $T(n)$ is no larger than a certain quantity, then it is no larger than a certain quantity plus something nonnegative e.g., $cn-n$ – Mike Jun 03 '18 at 22:06