Neural Network Optimization
Diana discusses the importance of characterizing neural networks in terms of latency and energy consumption to enhance hardware design. She emphasizes the need for methodologies that combine hardware metrics with accuracy, enabling better optimization of neural network architectures. By leveraging existing knowledge from the computer hardware domain, we can identify efficient configurations and take advantage of low-hanging fruit in performance modeling.In this clip
From this podcast
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
The Case for Hardware-ML Model Co-design with Diana Marculescu - #391
Related Questions