Methods › Sequential › Recurrent Neural Networks › QRNN

Quasi-Recurrent Neural Network

QRNN

15 papers tagged archive 2025-07-28

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

A QRNN, or Quasi-Recurrent Neural Network, is a type of recurrent neural network that alternates convolutional layers, which apply in parallel across timesteps, and a minimalist recurrent pooling function that applies in parallel across channels. Due to their increased parallelism, they can be up to 16 times faster at train and test time than LSTMs.

Given an input sequence 𝐗 ∈ℝ^(T×n) of T n-dimensional vectors 𝐱₁, …, 𝐱_T, the convolutional subcomponent of a QRNN performs convolutions in the timestep dimension with a bank of m filters, producing a sequence 𝐙 ∈ℝ^(T×m) of m-dimensional candidate vectors 𝐳ₜ. Masked convolutions are used so filters can not access information from future timesteps (implementing with left padding).

Additional convolutions are applied with separate filter banks to obtain sequences of vectors for the elementwise gates that are needed for the pooling function. While the candidate vectors are passed through a tanh nonlinearity, the gates use an elementwise sigmoid. If the pooling function requires a forget gate fₜ and an output gate oₜ at each timestep, the full set of computations in the convolutional component is then:

𝐙 = tanh(𝐖_z ∗ 𝐗) 𝐅 = σ(𝐖_f ∗ 𝐗) 𝐎 = σ(𝐖ₒ ∗ 𝐗)

where 𝐖_z, 𝐖_f, and 𝐖ₒ, each in ℝ^(k×n×m), are the convolutional filter banks and ∗ denotes a masked convolution along the timestep dimension. Dynamic average pooling by Balduzzi & Ghifary (2016) is used, which uses only a forget gate:

𝐡ₜ = 𝐟ₜ ⊙𝐡ₜ₋₁ + (1 − 𝐟ₜ) ⊙ 𝐳ₜ

Which is denoted f-pooling. The function may also include an output gate:

𝐜ₜ = 𝐟ₜ ⊙𝐜ₜ₋₁ + (1 − 𝐟ₜ) ⊙ 𝐳ₜ

𝐡ₜ = 𝐨ₜ ⊙𝐜ₜ

Which is denoted fo-pooling. Or the recurrence relation may include an independent input and forget gate:

𝐜ₜ = 𝐟ₜ ⊙𝐜ₜ₋₁ + 𝐢ₜ⊙ 𝐳ₜ

𝐡ₜ = 𝐨ₜ ⊙𝐜ₜ

Which is denoted ifo-pooling. In each case h or c is initialized to zero. The recurrent part sof these functions must be calculated for each timestep in the sequence, but parallelism along feature dimensions means evaluating them even over long sequences requires a negligible amount of computation time.

A single QRNN layer thus performs an input-dependent pooling, followed by a gated linear combination of convolutional features. As with convolutional neural networks, two or more QRNN layers should be stacked to create a model with the capacity to approximate more complex functions.

Source: Quasi-Recurrent Neural Networks

Papers archive 2025-07-28

15 shown of 15, 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 34 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
Time Series Prediction3
Decision Making2
Language Modeling2
Language Modelling2
Sentiment Analysis2
Sentiment Classification2
Spatial Interpolation2
Time Series2
Time Series Analysis2
quantile regression2
regression2
Benchmarking1
CPU1
Collision Avoidance1
Data Augmentation1
Ensemble Learning1
Feature Engineering1
Feature Importance1
General Classification1
Image Augmentation1

Usage over time archive 2025-07-28

Papers per year tagged with QRNN: 2016 to 2024, peak 5 5 0 2016: 1 paper 2016 2017: 1 paper 2017 2018: 1 paper 2018 2019: 0 papers 2019 2020: 2 papers 2020 2021: 1 paper 2021 2022: 2 papers 2022 2023: 5 papers 2023 2024: 2 papers 2024
Papers per year the archive tags with this method, by the paper's archive date (15 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