Methods › Natural Language Processing › Synthesized Attention Mechanisms › Factorized Random Synthesized Attention
Factorized Random Synthesized Attention
Introduced by Yi Tay et al. in Synthesizer: Rethinking Self-Attention in Transformer Models
archive 2025-07-28 Description, source and code snippet are the archive's method entry.
Factorized Random Synthesized Attention, introduced with the Synthesizer architecture, is similar to factorized dense synthesized attention but for random synthesizers. Letting R being a randomly initialized matrix, we factorize R into low rank matrices R₁, R₂ ∈ℝ^(l xk) in the attention function:
Y = Softmax(R₁R₂ᵀ)G(X) .
Here G(.) is a parameterized function that is equivalent to V in Scaled Dot-Product Attention.
For each head, the factorization reduces the parameter costs from l² to 2(lk) where k << l and hence helps prevent overfitting. In practice, we use a small value of k = 8.
The basic idea of a Random Synthesizer is to not rely on pairwise token interactions or any information from individual token but rather to learn a task-specific alignment that works well globally across many samples.
Papers archive 2025-07-28
1 shown of 1, newest first. Repository counts are the archive's code-links table. A Syntology line states what Syntology ran from that paper's harvested code; it is per sample and not a correctness claim.
-
Synthesizer: Rethinking Self-Attention in Transformer Models 2 May 2020 · 1 repository · arXiv:2005.00743Syntology ran 1 of 1 samples · 0 unverified
Tasks archive 2025-07-28
10 tasks the archive attaches to papers tagged with this method, by distinct papers. A task without a page in the catalog is plain text.
Usage over time archive 2025-07-28
Components: the archive holds no method-to-method composition, so PwC's Components table cannot be rebuilt; the Papers list carries no Results column for the same reason (the archive does not join its leaderboard rows to method tags).
Categories archive 2025-07-28
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections