Hybrid Real- and Complex-valued Neural Network Architecture
Alex Young NXP SemiconductorsHigh Tech Campus5656 AE Eindhoven, The Netherlands, Luan Vinícius Fiorio Department of Electrical EngineeringEindhoven University of Technology5600 MB Eindhoven, The Netherlands, Bo YangDepartment of Electrical EngineeringEindhoven University of Technology5600 MB Eindhov · 2025
Read Hybrid Real- and Complex-valued Neural Network Architecture by Alex Young NXP SemiconductorsHigh Tech Campus5656 AE Eindhoven, The Netherlands, Luan Vinícius Fiorio Department of Electrical EngineeringEindhoven University of Technology5600 MB Eindhoven, The Netherlands, Bo YangDepartment of Electrical EngineeringEindhoven University of Technology5600 MB Eindhov (2025) 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-2504-03497.txt.
Description
Alex Young and co-authors build neural networks that think partly in complex numbers in a 2025 paper. Many signals, from audio to radio, are naturally complex-valued, carrying phase as well as size. Real-valued networks can still learn such tasks, but the paper shows they reinvent complex arithmetic clumsily, wasting capacity to mimic what complex units do natively.
Their hybrid architecture mixes real and complex paths inside each block, with conversion functions passing information between the two worlds. New complex activation functions improve generalization and parameter thrift. A dedicated search procedure explores the enlarged design space. On a spoken-digit audio benchmark, the hybrid reached lower loss with fewer parameters than a purely real network in every configuration tried. The authors pitch partial complexity as a general tool for signal processing tasks.
Catalog record
| CREATOR | Alex Young NXP SemiconductorsHigh Tech Campus5656 AE Eindhoven, The Netherlands, Luan Vinícius Fiorio Department of Electrical EngineeringEindhoven University of Technology5600 MB Eindhoven, The Netherlands, Bo YangDepartment of Electrical EngineeringEindhoven University of Technology5600 MB Eindhov |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-03497