Sketching as a Tool for Understanding and Accelerating Self-attention for Long Sequences
Chen, Yifan, Zeng, Qi, Hakkani-Tur, Dilek, Jin, Di, Ji, Heng, Yang, Yun
Transformer-based models are not efficient in processing long sequences due to the quadratic space and time complexity of the self-attention modules. To address this limitation, Linformer and Informer are proposed to reduce the quadratic complexity to linear (modulo logarithmic factors) via low-dimensional projection and row selection respectively. These two models are intrinsically connected, and to understand their connection, we introduce a theoretical framework of matrix sketching. Based on the theoretical analysis, we propose Skeinformer to accelerate self-attention and further improve the accuracy of matrix approximation to self-attention with three carefully designed components: column sampling, adaptive row normalization and pilot sampling reutilization. Experiments on the Long Range Arena (LRA) benchmark demonstrate that our methods outperform alternatives with a consistently smaller time/space footprint.
Dec-10-2021
- Country:
- North America
- United States
- Oregon (0.04)
- Louisiana > Orleans Parish
- New Orleans (0.04)
- Illinois > Champaign County
- Urbana (0.04)
- California
- San Diego County > San Diego (0.04)
- Los Angeles County > Long Beach (0.04)
- Canada
- Quebec > Montreal (0.04)
- British Columbia > Metro Vancouver Regional District
- Vancouver (0.04)
- United States
- Europe
- France (0.04)
- United Kingdom > England
- Oxfordshire > Oxford (0.04)
- Cambridgeshire > Cambridge (0.04)
- Italy > Lazio
- Rome (0.04)
- Asia > China
- Hong Kong (0.04)
- Africa > Ethiopia
- Addis Ababa > Addis Ababa (0.04)
- North America
- Genre:
- Research Report (0.64)
- Industry:
- Technology: