MULTISCALE ATTENTION VIA WAVELET NEURAL OPERATORS FOR VISION TRANSFORMER
Primary Area: representation learning for computer vision, audio, language, and other modalities
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: vision transformers, Wavelet transform, neural operators, attention
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Abstract: Transformers have achieved widespread success in computer vision. At their heart, there is a self-attention mechanism, an inductive bias that associates each token in the input with every other token through a weighted basis. The standard self-attention has quadratic complexity with the sequence length, which impedes its utility to long sequences appearing in high resolution vision. Recently, inspired by operator learning for PDEs, adaptive Fourier neural operators (AFNO) were introduced for high resolution attention based on global convolution that is efficiently implemented via FFT. However, the AFNO global filtering cannot well represent small and moderate scale structures that commonly appear in natural images. To leverage the coarse-to-fine scale structures we introduce a multiscale Wavelet attention (MWA) by leveraging wavelet neural operators which incurs linear complexity in the sequence size. We replace the attention in ViT with MWA and our experiments with CIFAR and Tiny-ImageNet classification demonstrate significant improvement over alternative Fourier-based attentions such as AFNO and global filter network (GFN).
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 135
Loading