Structured State Space Sequence Model (S4)
Definition
S4 models long sequences with a structured state space model.
A concrete sequence model architecture that makes a structured parametrized linear state space model computable so that training can be done as convolution and inference recurrently.
Simply explained
An internal state describes the progression of a long sequence and is advanced with new inputs.
Why it matters in practice
The term helps to select, develop, evaluate AI systems in a well-founded manner and realistically assess their technical risks.
Example from business
An S4 model processes long sensor signals with convolutional training and state-based inference.
Limits and pitfalls
- Using S4 as a generic term for all state-space models, equating linear state dynamics with arbitrary nonlinearity, or comparing efficiency claims without sequence length and hardware context.
Origin & history
Gu, Goel, and Ré published S4 in 2021.
Use in mid-sized companies
- Possible application: An S4 model processes long sensor signals with convolutional training and state-based inference.
Frequently asked questions
What does Structured State Space Sequence Model (S4) mean?
A concrete sequence model architecture that makes a structured parametrized linear state space model computationally feasible so that training can be done as convolution and inference can be recurrent. What mistakes should be avoided? Using S4 as a generic term for all state-space models, equating linear state dynamics with arbitrary nonlinearity, or comparing efficiency claims without considering sequence length and hardware.
Sources
Source checked on 10.10.2026. Editorial application examples are hypothetical; no evidence of real customer results.
Editorially reviewed on: