Sequential Reservoir Computing for Efficient High-Dimensional Spatiotemporal Forecasting
Ata Akbari Asanjan, Filip Wudarski, Daniel O’Connor, Shaun Geaney, Elena Strbac, P. Aaron Lott, Davide Venturelli · 2026
Read Sequential Reservoir Computing for Efficient High-Dimensional Spatiotemporal Forecasting by Ata Akbari Asanjan, Filip Wudarski, Daniel O’Connor, Shaun Geaney, Elena Strbac, P. Aaron Lott, Davide Venturelli (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-00172.txt.
Description
Ata Akbari Asanjan leads six coauthors in speeding up forecasts of systems that vary in space and time. Recurrent networks train slowly on such data, and memory becomes a bottleneck. Reservoir computing avoids backpropagation by fixing the recurrent layer. Only a simple readout is trained. Yet a single large reservoir still scales badly with input size.
Their design splits one reservoir into a chain of smaller ones. The linked chain keeps long term memory at lower cost. Tests cover a small chaotic system and two fluid simulations. Forecast horizons grow by a fifth or more. Error scores drop by a similar share. Training cost falls by up to a thousandfold against recurrent baselines. The authors offer the chain as a practical route. Real time, low energy forecasting comes within easier reach.
Catalog record
| CREATOR | Ata Akbari Asanjan, Filip Wudarski, Daniel O’Connor, Shaun Geaney, Elena Strbac, P. Aaron Lott, Davide Venturelli |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-00172