Скачать с ютуб Constrained Zero-Shot Neural Architecture Search on Small Classification Dataset в хорошем качестве

Constrained Zero-Shot Neural Architecture Search on Small Classification Dataset

The rapid evolution of Deep Learning has revolutionized scientific domains, yet it still remains a challenge for edge applications that impose strict constraints on memory and energy. In this talk, we highlight these challenges and present method based on zero-shot Neural Architecture Search (NAS) to automate the design of networks for edge devices. Our method achieves improved accuracy and reduced memory footprint, validated on an image classification problem, and enables efficient on-device processing. Speaker: Rémy Vuagniaux Associate R&D Engineer CSEM