Hardware aware neural search speeds up chip specific design optimization
MOSHO: A Native Discrete Shark-Inspired Metaheuristic for Hardware-Aware Neural Architecture Search
Neural and Evolutionary ComputingArtificial IntelligenceMachine Learning
Summary
Designing computer brains called neural networks for specific devices requires balancing accuracy and speed. The authors show that using a new method inspired by shark hunting that works directly with the building blocks of these networks helps find the best designs faster. Their approach avoids errors from previous methods and finds good solutions much more quickly with almost the same quality. This can save time and computing resources when customizing AI for devices like phones or sensors.
What this means in practice
- •For hardware design teams: Optimize neural network architectures to meet specific performance and speed targets for different hardware devices efficiently.
- •For edge device developers: Reduce development time to find neural network designs that run quickly and accurately on constrained devices like smartphones and IoT sensors.
Authors
Nazim Abdeerhmane Aouni, Zakaria Soualah Mohammed, Maroua Ogab, Nour El Imane Elbar, Rayane Rahmat Errahmane Smara, Billel Moussous, Adriane Anis Khaled, Malika Bessedik, Fatima Benbouzid Si Tayeb, Hadjer Benmeziane
Abstract
Hardware-aware neural architecture search (NAS) requires jointly optimizing predictive accuracy and device-specific latency over discrete search spaces. While multi-objective swarm optimizers show promise, continuous formulations such as the Multi-Objective White Shark Optimizer (MOWSO) can suffer from quantization noise when rounded to integer-encoded architecture vectors. To address this limitation, we propose MOSHO (Multi-Objective Shark Hunting Optimization), a native discrete metaheuristic that replaces continuous motion equations with discrete architecture-editing operators. MOSHO integrates adaptive operator scheduling, energy-regulated restarts, and deferred crowding-distance truncation to enhance exploration and search efficiency. On the merged NAS-Bench-201/HW-NAS-Bench benchmark across 18 contexts, MOSHO achieves a median normalized hypervolume of 0.9987 and IGD+ of 0.0064, yielding Pareto fronts competitive with NSGA-II while reducing search runtime by 96.4% (0.64 s vs. 17.72 s). These results demonstrate that tailored discrete search mechanisms can enable swarm optimizers to achieve high-quality multi-objective solution quality with substantially lower search cost in hardware-aware NAS.