AI Science Tracker
EN

Physics-guided learning reaches a measured radio-link threshold with fewer samples

University of Warwick · 20 Oct 2026 · physics

Discovery

Physics-guided learning reaches a measured radio-link threshold with fewer samples

Source check: Source checked. Not independent scientific confirmation.

Finding: A shallow physics-guided network fitted a radio-link model from received-signal-strength measurements for a fixed reconfigurable surface. In the tested setup it reached53dB with400 measurements, versus1600 for the compared CSM method. This is a setup-specific recorded-measurement validation, not a general mobile-network or energy-efficiency benefit.

Technology: Parametric physics-guided neural blind beamforming; P-PGNBB

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
The fitted physics-guided network estimates a configuration-to-signal mapping used to select surface settings, enabling the measured sample-budget comparison.
Test and result
A14-parameter model learned from measured surface configurations and signal strengths. A256-element surface and fixed transmitter/receiver geometry supplied the field dataset.
Limitation / still unknown
The comparison concerns one fixed setup and subsets of a recorded acquisition. Test prediction error was about4.38dB; noise, model mismatch and temporal variation were not separately isolated. More flexible PGNN had lower prediction error at larger budgets. Classical phase-coherence observations and perfect-CSI simulation bounds are separate from the learned result.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2610.27972v1 · https://arxiv.org/html/2610.27972v1

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 21 Nov 2026. Review scope: Original v1 front, actual learned model Methods, clinical/physical-reference evaluation and limitations.

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 1/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open in the full catalog for filters, charts and related discoveries. Browse physics discoveries.

Related catalog entries

Links use documented relationships where available, plus shared catalog context. A shared topic or model family does not establish a scientific dependency.