Pay Attention to MLPs
This challenges the necessity of self-attention in Transformers, offering a simpler alternative for researchers and practitioners in NLP and computer vision, though it is incremental in proposing a new architecture.
The paper tackles the problem of Transformer reliance in deep learning by proposing gMLP, a simple MLP-based architecture with gating, showing it performs as well as Transformers in language and vision tasks, achieving parity in accuracy and perplexity, and scaling similarly with data and compute.
Transformers have become one of the most important architectural innovations in deep learning and have enabled many breakthroughs over the past few years. Here we propose a simple network architecture, gMLP, based on MLPs with gating, and show that it can perform as well as Transformers in key language and vision applications. Our comparisons show that self-attention is not critical for Vision Transformers, as gMLP can achieve the same accuracy. For BERT, our model achieves parity with Transformers on pretraining perplexity and is better on some downstream NLP tasks. On finetuning tasks where gMLP performs worse, making the gMLP model substantially larger can close the gap with Transformers. In general, our experiments show that gMLP can scale as well as Transformers over increased data and compute.