On 12/08/2015 04:55 PM, Timon Gehr wrote:
The context is asymptotic runtime bounds. The special cases for small values of continuous logarithms can just be defined away. They have no relevance for asymptotic runtime analysis (even though defining big-O adequately for multiple variables is more tricky than for a single variable).
From the cycle "Andrei's esprits d'escalier", the inequality can be derived from the somewhat notorious log sum inequality if you make all bi equal: http://ocw.usu.edu/Electrical_and_Computer_Engineering/Information_Theory/lecture3.pdf
Andrei
