Methods › Sequential › Recurrent Neural Networks › SRU

SRU

16 papers tagged archive 2025-07-28

Introduced by Tao Lei et al. in Simple Recurrent Units for Highly Parallelizable Recurrence

archive 2025-07-28 Description, source and code snippet are the archive's method entry.

SRU, or Simple Recurrent Unit, is a recurrent neural unit with a light form of recurrence. SRU exhibits the same level of parallelism as convolution and feed-forward nets. This is achieved by balancing sequential dependence and independence: while the state computation of SRU is time-dependent, each state dimension is independent. This simplification enables CUDA-level optimizations that parallelize the computation across hidden dimensions and time steps, effectively using the full capacity of modern GPUs.

SRU also replaces the use of convolutions (i.e., ngram filters), as in QRNN and KNN, with more recurrent connections. This retains modeling capacity, while using less computation (and hyper-parameters). Additionally, SRU improves the training of deep recurrent models by employing highway connections and a parameter initialization scheme tailored for gradient propagation in deep architectures.

A single layer of SRU involves the following computation:

𝐟ₜ =σ(𝐖_f 𝐱ₜ+𝐯_f ⊙𝐜ₜ₋₁+𝐛_f)

𝐜ₜ =𝐟ₜ ⊙𝐜ₜ₋₁+(1-𝐟ₜ) ⊙(𝐖 𝐱ₜ)

𝐫ₜ =σ(𝐖ᵣ 𝐱ₜ+𝐯ᵣ ⊙𝐜ₜ₋₁+𝐛ᵣ)

𝐡ₜ =𝐫ₜ ⊙𝐜ₜ+(1-𝐫ₜ) ⊙𝐱ₜ

where 𝐖, 𝐖_f and 𝐖ᵣ are parameter matrices and 𝐯_f, 𝐯ᵣ, 𝐛_f and 𝐛ᵥ are parameter vectors to be learnt during training. The complete architecture decomposes to two sub-components: a light recurrence and a highway network,

The light recurrence component successively reads the input vectors 𝐱ₜ and computes the sequence of states 𝐜ₜ capturing sequential information. The computation resembles other recurrent networks such as LSTM, GRU and RAN. Specifically, a forget gate 𝐟ₜ controls the information flow and the state vector 𝐜ₜ is determined by adaptively averaging the previous state 𝐜ₜ₋₁ and the current observation 𝐖 𝐱₊according to 𝐟ₜ.

PaperSource

Papers archive 2025-07-28

16 shown of 16, 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.

Tasks archive 2025-07-28

20 shown of 27 tasks the archive attaches to papers tagged with this method, by distinct papers. A task without a page in the catalog is plain text.

TaskPapers
Speech Recognition4
speech-recognition4
Translation3
Data Augmentation2
General Classification2
Language Modeling2
Language Modelling2
Machine Translation2
Question Answering2
CPU1
Denoising1
Diversity1
NMT1
Quantization1
Reading Comprehension1
Retrieval1
Robot Navigation1
Session-Based Recommendations1
Speech Denoising1
Speech Enhancement1

Usage over time archive 2025-07-28

Papers per year tagged with SRU: 2017 to 2024, peak 5 5 0 2017: 1 paper 2017 2018: 2 papers 2018 2019: 2 papers 2019 2020: 5 papers 2020 2021: 1 paper 2021 2022: 1 paper 2022 2023: 3 papers 2023 2024: 1 paper 2024
Papers per year the archive tags with this method, by the paper's archive date (16 dated). Bars are counts, not a trend claim.

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

Recurrent Neural Networks

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