Saved in:
Bibliographic Details
Main Authors: Cunningham, Harry Jake, Giannone, Giorgio, Zhang, Mingtian, Deisenroth, Marc Peter
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2408.09453
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911994321305600
author Cunningham, Harry Jake
Giannone, Giorgio
Zhang, Mingtian
Deisenroth, Marc Peter
author_facet Cunningham, Harry Jake
Giannone, Giorgio
Zhang, Mingtian
Deisenroth, Marc Peter
contents Global convolutions have shown increasing promise as powerful general-purpose sequence models. However, training long convolutions is challenging, and kernel parameterizations must be able to learn long-range dependencies without overfitting. This work introduces reparameterized multi-resolution convolutions ($\texttt{MRConv}$), a novel approach to parameterizing global convolutional kernels for long-sequence modelling. By leveraging multi-resolution convolutions, incorporating structural reparameterization and introducing learnable kernel decay, $\texttt{MRConv}$ learns expressive long-range kernels that perform well across various data modalities. Our experiments demonstrate state-of-the-art performance on the Long Range Arena, Sequential CIFAR, and Speech Commands tasks among convolution models and linear-time transformers. Moreover, we report improved performance on ImageNet classification by replacing 2D convolutions with 1D $\texttt{MRConv}$ layers.
format Preprint
id arxiv_https___arxiv_org_abs_2408_09453
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Reparameterized Multi-Resolution Convolutions for Long Sequence Modelling
Cunningham, Harry Jake
Giannone, Giorgio
Zhang, Mingtian
Deisenroth, Marc Peter
Machine Learning
Computer Vision and Pattern Recognition
Global convolutions have shown increasing promise as powerful general-purpose sequence models. However, training long convolutions is challenging, and kernel parameterizations must be able to learn long-range dependencies without overfitting. This work introduces reparameterized multi-resolution convolutions ($\texttt{MRConv}$), a novel approach to parameterizing global convolutional kernels for long-sequence modelling. By leveraging multi-resolution convolutions, incorporating structural reparameterization and introducing learnable kernel decay, $\texttt{MRConv}$ learns expressive long-range kernels that perform well across various data modalities. Our experiments demonstrate state-of-the-art performance on the Long Range Arena, Sequential CIFAR, and Speech Commands tasks among convolution models and linear-time transformers. Moreover, we report improved performance on ImageNet classification by replacing 2D convolutions with 1D $\texttt{MRConv}$ layers.
title Reparameterized Multi-Resolution Convolutions for Long Sequence Modelling
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2408.09453