Proceedings of the Web Conference 2020 2020
DOI: 10.1145/3366423.3380160
|View full text |Cite
|
Sign up to set email alerts
|

Fast Generating A Large Number of Gumbel-Max Variables

Abstract: The well-known Gumbel-Max Trick for sampling elements from a categorical distribution (or more generally a nonnegative vector) and its variants have been widely used in areas such as machine learning and information retrieval. To sample a random element i (or a Gumbel-Max variable i) in proportion to its positive weight v i , the Gumbel-Max Trick first computes a Gumbel random variable д i for each positive weight element i, and then samples the element i with the largest value of д i + ln v i . Recently, appl… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1

Citation Types

0
3
0

Year Published

2021
2021
2024
2024

Publication Types

Select...
3
2

Relationship

1
4

Authors

Journals

citations
Cited by 5 publications
(3 citation statements)
references
References 42 publications
0
3
0
Order By: Relevance
“…Therefore, the same sampling process may give rise to different distributions, depending on what is considered 'the sample'. When repeated samples from the categorical are drawn using the Gumbel-max trick, it may be possible to gain efficiency by reusing computations, as proposed by the authors of [107], [125].…”
Section: Unstructured Distributionsmentioning
confidence: 99%
“…Therefore, the same sampling process may give rise to different distributions, depending on what is considered 'the sample'. When repeated samples from the categorical are drawn using the Gumbel-max trick, it may be possible to gain efficiency by reusing computations, as proposed by the authors of [107], [125].…”
Section: Unstructured Distributionsmentioning
confidence: 99%
“…They also demonstrated that the probability Jaccard similarity is scale invariant and more sensitive to changes in vectors. Qi et al [40] proposed FastGM to further reduce the time complexity of P-MinHash by generating the hash values in order.…”
Section: Related Workmentioning
confidence: 99%
“…When repeated samples from the categorical are drawn using the Gumbel-max trick, it may be possible to gain efficiency by reusing computations, as proposed by the authors of [104], [122].…”
Section: Structured Distributionmentioning
confidence: 99%