Hybrid Dealiased Convolutions

2023/05/14 by George, Robert Joseph, Murasko, Noel, Bowman, John C.
#FOS: Mathematics #Numerical Analysis (math.NA)

paper · doi:10.48550/arxiv.2306.10016

Abstract

This paper proposes a practical and efficient solution for computing convolutions using hybrid dealiasing. It offers an alternative to explicit or implicit dealiasing and includes an optimized hyperparameter tuning algorithm that uses experience to find the optimal parameters. Machine learning algorithms and efficient heuristics are also developed to estimate optimal parameters for larger convolution problems using only small squares/rectangles.

Related

• Hybrid Dealiasing of Complex Convolutions
• Efficient Dealiased Convolutions without Padding
• Faster Convolution Inference Through Using Pre-Calculated Lookup Tables
• Efficient Learning of Convolution Weights as Gaussian Mixture Model Posteriors
• Generalized Convolution and Efficient Language Recognition