Deep sequence models develop geometric memory in embeddings that encodes novel global relationships, transforming l-fold composition tasks into 1-step navigation via a natural spectral bias connected to Node2Vec.
Linear transformers are secretly fast weight programmers
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
citation-role summary
other 1
citation-polarity summary
fields
cs.LG 2verdicts
UNVERDICTED 2roles
other 1polarities
unclear 1representative citing papers
Ordinary least squares is a special case of the single-layer linear transformer when attention parameters are set via spectral decomposition of the empirical covariance matrix.
citing papers explorer
-
Deep sequence models tend to memorize geometrically; it is unclear why
Deep sequence models develop geometric memory in embeddings that encodes novel global relationships, transforming l-fold composition tasks into 1-step navigation via a natural spectral bias connected to Node2Vec.
-
Ordinary Least Squares is a Special Case of Transformer
Ordinary least squares is a special case of the single-layer linear transformer when attention parameters are set via spectral decomposition of the empirical covariance matrix.