Signal Propagation in Transformers: Theoretical Perspectives and the Role of Rank Collapse

Item #:

068431-1972

Details

DOI: https://doi.org/10.52202/068431-1972
Author(s): Sotiris Anagnostidis, Luca Biggio, Aurelien Lucchi, Lorenzo Noci, Antonio Orvieto, Sidak Pal Singh
Pages: 27198-27211 (14 pages)
Format: PDF Paper Download
Conference: Advances in Neural Information Processing Systems 35
Date/Location: Held 28 November - 9 December 2022, New Orleans, Louisiana, USA.
Series: Advances in Neural Information Processing Systems 35
Publisher: Neural Information Processing Systems Foundation, Inc. (NeurIPS)

DOI: https://doi.org/10.52202/068431-1972
Author(s): Sotiris Anagnostidis, Luca Biggio, Aurelien Lucchi, Lorenzo Noci, Antonio Orvieto, Sidak Pal Singh
Pages: 27198-27211 (14 pages)
Format: PDF Paper Download
Conference: Advances in Neural Information Processing Systems 35
Date/Location: Held 28 November - 9 December 2022, New Orleans, Louisiana, USA.
Series: Advances in Neural Information Processing Systems 35
Publisher: Neural Information Processing Systems Foundation, Inc. (NeurIPS)