Successive Halving Top-k Operator
Tytuł:
Successive Halving Top-k Operator
Konferencja:
National Conference of the American Association for Artificial Intelligence [AAAI]
Rok:
2021
Opis:
We propose a differentiable successive halving method of relaxing the top-k operator, rendering gradient-based optimization possible. The need to perform softmax iteratively on the entire vector of scores is avoided by using a tournament-style selection. As a result, a much better approximation of top-k with lower computational cost is achieved compared to the previous approach.
Link:
https://arxiv.org/abs/2010.15552