Complex-to-real sketches for tensor products with applications to the polynomial kernel

Wacker, Jonas; Ohana, Ruben; Filippone, Maurizio
AISTATS 2023, 26th Annual International Conference on Artificial Intelligence and Statistics, 25-27 April 2023, València, Spain

Randomized sketches of a tensor product of p vectors follow a tradeoff between statistical efficiency and computational acceleration. Commonly used approaches avoid computing the high-dimensional tensor product explicitly, resulting in a suboptimal dependence of O(3p ) in the embedding dimension. We propose a simple Complex-to-Real (CtR) modification of wellknown sketches that replaces real random projections by complex ones, incurring a lower O(2p ) factor in the embedding dimension. The output of our sketches is real-valued, which renders their downstream use straightforward. In particular, we apply our sketches to p-fold self-tensored inputs corresponding to the feature maps of the polynomial kernel. We show that our method achieves state-of-the-art performance in terms of accuracy and speed compared to other randomized approximations from the literature. 


HAL
Type:
Poster / Demo
City:
Valencia
Date:
2023-04-25
Department:
Data Science
Eurecom Ref:
7276

PERMALINK : https://www.eurecom.fr/publication/7276