HERCULES: Hardware-Efficient, Robust, Continual Learning Neural Architecture Search
Matteo Gambella, Fabrizio Pittorino, and Manuel Roveri · 2026
Read HERCULES: Hardware-Efficient, Robust, Continual Learning Neural Architecture Search by Matteo Gambella, Fabrizio Pittorino, and Manuel Roveri (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-2605-04103.txt.
Description
Matteo Gambella, Fabrizio Pittorino, and Manuel Roveri survey neural architecture search through a triple lens. They argue modern practice now requires three things at once: efficiency, robustness, and continual learning. The search finds network designs that balance accuracy and cost on its own. But as AI moves to real-world deployment, hardware-aware efficiency alone is no longer sufficient. That is especially true at the edge. Robustness covers environmental variability. Continual learning covers adaptation without forgetting.
The authors propose a taxonomy. It sorts methods by aim: resource optimization, environmental resilience, and architectural plasticity. They argue the three axes reinforce each other rather than compete. Building on it, they map the landscape into a framework named HERCULES. They define its wish list, styled as twelve labours, for search that is hardware-efficient, robust, and continual. The survey is aimed at people who build such systems.
Catalog record
| CREATOR | Matteo Gambella, Fabrizio Pittorino, and Manuel Roveri |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 95 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.04103