EdgeVolution: democratizing multi-objective neural architecture search and end-to-end deployment on microcontrollers.

Inference time, particularly for the best-performing models in each generation, demonstrates a decreasing trend throughout the optimization process (Fig. 4 c).

also in 47,180 other papers

To address this challenge, there is an increasing trend to combine neural architecture search (NAS) with hardware-aware optimization strategies, formulating it as a multi-objective optimization problem that balances model performance against hardware constraints 10 , 11 .

also in 63,971 other papers