+1 vote

Given f(n) = o(g(n)), show that it is not necessary that  log (f(n)) = o(log (g(n)))

[Hint, use a counterexample]

in Asymptotic Analysis by AlgoMeister (1.6k points)

2 Answers

+3 votes
 
Best answer

A simple contradiction would be nand n3.

While n= o (n3),

Log(n2) = 2 Log n,  Log(n3) = 3 Log n , Hence Log(n2) = θ ( Log(n3))

by AlgoMeister (696 points)
selected by
–1 vote

lim(n->infinity)(log(f(n))/log(g(n))) = lim n->infinity (1/(f(n)*ln2) * f'(n)) / (1/(g(n) * ln2)) * g'(n))

                                                     = lim n->infinity (g(n)/f(n) * f'(n)/g'(n))

lim n->infinity (g(n)/f(n)) = infinity

lim n->infinity (f'(n)/g'(n)) = 0

let f(n) = n, g(n) = n^2

lim n-> infinity (f(n)/g(n)) =  lim n-> infinity (1/n)

                                       = 0

f(n) = o(g(n))​

In this condition

lim(n->infinity)(log(f(n))/log(g(n)))  = lim n->infinity (n * 1/2n)

                                                      = 1

so it is not necessary that  log (f(n)) = o(log (g(n))) when f(n) = o(g(n))

by (124 points)
I don't know what the first 3 lines add to this proof by counter example.  Wouldn't you just want to start with details of f(n) and g(n)?

Related questions

0 votes
2 answers
+3 votes
1 answer
The Book: Analysis and Design of Algorithms | Presentations on Slideshare | Lecture Notes, etc
...