Discover and read the best of Twitter Threads about #nocollision
Most recents (1)
📊 Two days ago, I asked a question. Way more people answered than expected, and... well, this week's weₐekly #quiz will be slightly ≠: a long thread on uniformity testing, trickling down all day long.
Be careful what you wish for :) #statistics
1/n
Be careful what you wish for :) #statistics
1/n
So.... uniformity testing. You have n i.i.d. samples from some unknown distribution over [k]={1,2,...,k} and want to know: is it *the* uniform distribution? Or is it statistically far from it, say, at total variation distance ε?
2/n
2/n
So, before delving deeper, let's recall what total variation distance is, and formalize the question. The former is basically a measure of how distinguishable two distributions are given a single sample.
en.wikipedia.org/wiki/Total_var…
The latter... is below.
3/n
en.wikipedia.org/wiki/Total_var…
The latter... is below.
3/n
