University of Cambridge > Talks.cam > Machine Learning Reading Group @ CUED > Linear Attention for Efficient Transformers

  Browse A-Z

Log in

University Account External (via Google)

🔐 Log In

Information on

Subscribing to talksdetails Finding a talkdetails Adding a talkdetails Disseminating talksdetails Help and Documentationdetails

ℹī¸ Information

Linear Attention for Efficient Transformers

Download to your calendar using vCal

If you have a question about this talk, please contact Xianda Sun .

Attention may be all you need, but that doesn’t mean it comes cheap. The Achilles’ Heel of the wildly successful Transformer architecture is its quadratic time- and space-complexity scaling with respect to the length of the input token sequence. A diverse taxonomy of methods has been proposed to remedy this bottleneck and recover linear complexity, including making attention local, sparse or low rank. We will explore the respective strengths and weaknesses of these approaches, discuss theoretical guarantees (or the lack thereof), and consider possible directions for future work.

Suggested reading:
  1. Attention is all you need (https://arxiv.org/abs/1706.03762). Seminal Transformers paper.
  2. Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention (https://arxiv.org/abs/2006.16236). Among the first papers on low-rank attention.
  3. Swin Transformer: Hierarchical Vision Transformer using Shifted Windows (https://arxiv.org/abs/2103.14030). Popular example of local attention.
  4. Big Bird: Transformers for Longer Sequences (https://arxiv.org/abs/2007.14062). Example of the benefits of using a combination of techniques.

This talk is part of the Machine Learning Reading Group @ CUED series.

This talk is included in these lists:

Note that ex-directory lists are not shown.

Other lists

POLIS MedImmune, Sir Aaron Klug Building, Granta Park, Cambridge, CB21 6GH In Situ Graduate Colloquium 2013 - Department of Architecture

Other talks

External Seminar - Michael Raissig TBC Title to be confirmed Title TBC Spectral theory for Dirac operators with singular potentials Title TBC Haematology and Palliative Care
 

Š 2006-2025 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity