Sliding-window training on readout layer enhances liquid state machines for temporal prediction
Shimeng Wu, Andrew Walter, Andy M. Tyrrell, Martin A. Trefzer · 2026
Read Sliding-window training on readout layer enhances liquid state machines for temporal prediction by Shimeng Wu, Andrew Walter, Andy M. Tyrrell, Martin A. Trefzer (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/pmc-13581606.txt.
Description
Shimeng Wu and co-authors improve liquid state machines for time prediction in a 2026 paper. These spiking reservoir systems suit low power, online forecasting, since only a simple readout layer is trained. Standard training reads a single instant of reservoir state, which misses evidence spread over time.
Feeding the readout a sliding window of recent states changed results sharply. On a standard benchmark, error for a ridge readout fell by about eighty five percent as window length grew, and a transformer readout improved by a similar share. The same trend held on two harder tasks. The lesson is that recent reservoir history, shown at the decoder, lifts cheap temporal prediction.
Catalog record
| CREATOR | Shimeng Wu, Andrew Walter, Andy M. Tyrrell, Martin A. Trefzer |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |