Tobias Nauen
PhD · efficient deep learning
01
Home
02
Publications
03
Software
04
Projects
05
Contact
⌘K
Home
/
taylor-shift
TaylorShift: Shifting the Complexity of Self-Attention from Squared to Linear (and Back) using Taylor-Softmax
ICPR 2024 (oral) · December 2024
Slides
Pdf
Code
Summary —
Oral presentation at ICPR 2024 introducing TaylorShift, a novel reformulation of the attention mechanism using Taylor-Softmax that enables full token-to-token interactions in linear time.
BibTeX
×
Copy
Download .bib