On Wed, Jul 31, 2013 at 5:19 PM, Anastasios Tsiolakidis <[email protected]> wrote: > On 31.07.2013, at 22:26, Matt Mahoney <[email protected]> wrote: > >> That is incorrect. Apple trees and water holes have a power law >> distribution, not a normal (Gaussian) distribution > > Oops, the Gaussian can't be found in nature, my bad lol!
Gaussian distributions are found whenever you add up a lot of random variables each with small variances compared to the sum. But there are a lot of cases where that isn't so. The distribution of symbols output by random programs has a distribution more like 1/t, where t is the time since it was last output regardless of what happened before. Many practical data compression programs use this model. For example, LZ77 (like zip) codes the symbol as a pointer to the previous occurrence using log(t) bits to code the distance. Since the optimal coding of a symbol with probability p is -log(p) bits, this implies a probability of 1/t. -- -- Matt Mahoney, [email protected] ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
