Masked Language Modeling for Proteins via Linearly Scalable Long-Context Transformers
arXiv:2006.03555
Abstract
Transformer models have achieved state-of-the-art results across a diverse range of domains. However, concern over the cost of training the attention mechanism to learn complex dependencies between distant inputs continues to grow. In response, solutions that exploit the structure and sparsity of the learned attention matrix have blossomed. However, real-world applications that involve long sequences, such as biological sequence analysis, may fall short of meeting these assumptions, precluding exploration of these models. To address this challenge, we present a new Transformer architecture, Performer, based on Fast Attention Via Orthogonal Random features (FAVOR). Our mechanism scales linearly rather than quadratically in the number of tokens in the sequence, is characterized by sub-quadratic space complexity and does not incorporate any sparsity pattern priors. Furthermore, it provides strong theoretical guarantees: unbiased estimation of the attention matrix and uniform convergence. It is also backwards-compatible with pre-trained regular Transformers. We demonstrate its effectiveness on the challenging task of protein sequence modeling and provide detailed theoretical analysis.
This arXiv submission has been deprecated. Please see "Rethinking Attention with Performers" at arXiv:2009.14794 for the most updated version of the paper
References in corpus (6)
- Identification of direct residue contacts in protein-protein interaction by message passing
- Generating Long Sequences with Sparse Transformers
- Conformer: Convolution-augmented Transformer for Speech Recognition
- Reformer: The Efficient Transformer
- Orthogonal Random Features
- Orthogonal Estimation of Wasserstein Distances
Cited by in corpus (7)
- Long Range Arena: A Benchmark for Efficient Transformers
- Fastformer: Additive Attention Can Be All You Need
- Learnable Fourier Features for Multi-Dimensional Spatial Positional Encoding
- Not All Memories are Created Equal: Learning to Forget by Expiring
- FNetAR: Mixing Tokens with Autoregressive Fourier Transforms
- H-Transformer-1D: Fast One-Dimensional Hierarchical Attention for Sequences
- THG: Transformer with Hyperbolic Geometry