-1

I am reading a natural language processing paper and I came across this expression. I don't know what it means. Especially the unif part.

$$m_i \sim \operatorname{unif}\{1,n\}\text{ for } i = 1 \text{ to } k$$

Michael Hardy
  • 1
odbhut.shei.chhele
  • 351
  • 1
    Well, it very well may mean the "uniform distribution." Hard to tell from what's here, but if the distrubution is discrete, each of these are equally probable: ${ 1,2,3,\cdots, n}$. – mjw Aug 05 '20 at 02:24
  • 2
    Since you are asking about the meaning of an expression found in "a natural language processing paper", it is essential that you give a citation of that paper. While one may suspect the notation describes a "uniform distribution" on a set, the paper itself will be best evidence of what the author intended. – hardmath Aug 05 '20 at 02:33
  • https://arxiv.org/abs/2003.10555 – odbhut.shei.chhele Aug 05 '20 at 04:10

1 Answers1

2

It means you have $k$ random variables, $m_1, m_2, \ldots , m_k$, each of which was chosen from a uniform distribution with extreme values $1$ and $n$. It is fundamentally ambiguous as to whether the distribution is discrete or continuous. (Check your source.)

Example: if $k = 3$ and $n = 5$, you could have...

...in the continuous case:

$m_1 = 2.938, m_2 = 1.155, m_3 = 4.076$.

... and in discrete case:

$m_1 = 4, m_2 = 5, m_3 = 1$.

Given your source is about natural language processing having (discrete) words, I suspect (but don't know) that the latter is the case.

David G. Stork
  • 29,774