CONF Mai_ACL2023_2023/IDIAP HyperMixer: An MLP-based Low Cost Alternative to Transformers Mai, Florian Pannatier, Arnaud Fehr, Fabio Chen, Haolin Marelli, François Fleuret, Francois Henderson, James EXTERNAL https://publications.idiap.ch/attachments/papers/2023/Mai_ACL2023_2023.pdf PUBLIC Association for Computational Linguistics - Proc. of the 61st Annual Meeting of the Association for Computational Linguistics Toronto, Canada 2023 15632-15654 978-1-959429-72-2 http://dx.doi.org/10.18653/v1/2023.acl-long.871 doi Transformer-based architectures are the model of choice for natural language understanding, but they come at a significant cost, as they have quadratic complexity in the input length, require a lot of training data, and can be difficult to tune. In the pursuit of lower costs, we investigate simple MLP-based architectures. We find that existing architectures such as MLPMixer, which achieves token mixing through a static MLP applied to each feature independently, are too detached from the inductive biases required for natural language understanding. In this paper, we propose a simple variant, HyperMixer, which forms the token mixing MLP dynamically using hypernetworks. Empirically, we demonstrate that our model performs better than alternative MLP-based models, and on par with Transformers. In contrast to Transformers, HyperMixer achieves these results at substantially lower costs in terms of processing time, training data, and hyperparameter tuning.