arXiv:2605.26600v1 Announce Type: new Abstract: Standard Self-Supervised Learning (SSL) for Automatic Modulation Recognition (AMR) struggles with ineffective isotropic augmentations, spectral instability, and semantic drift. To address these challenges, we propose Dynamic-Consistency Contrastive Learning (DyCo-CL), a geometry-aware framework that couples Virtual Adversarial Augmentation (VAA) with a semantic consistency loss. We provide a theoretical analysis indicating that this strategy acts as an implicit spectral regularizer for the encoder, enabling stable manifold exploration. Complement
The increasing complexity and practical deployment demands of AI systems, particularly in sensitive areas like automatic modulation recognition, necessitate more robust and reliable learning methods.
Improving the robustness and efficiency of few-shot learning in areas like Automatic Modulation Recognition directly impacts the performance of advanced communication and defense systems, reducing errors and increasing reliability.
This research introduces a more stable and geometry-aware method for self-supervised learning, potentially leading to more accurate and reliable AI deployments in critical applications, especially when data is scarce.
- · Defense contractors
- · Telecommunications companies
- · AI research labs
- · Signal processing engineers
- · Systems highly reliant on less robust SSL methods
More resilient AI models for signal processing will emerge, reducing false positives/negatives in critical systems.
The improved robustness could accelerate the deployment of autonomous systems in complex electromagnetic environments.
This could lead to a competitive advantage for nations or entities that successfully integrate such advanced AI into their defense and communication infrastructure.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG