DNN Workload

DNN workload research focuses on optimizing the execution of deep neural networks (DNNs), encompassing single and multi-DNN scenarios, across diverse hardware platforms ranging from mobile devices to large data centers. Current efforts concentrate on improving efficiency through techniques like model sparsification, novel scheduling algorithms (e.g., dynamic and static scheduling), and hardware acceleration using specialized architectures (e.g., spiking neural networks, optical computing, and tensor cores). These advancements aim to address critical bottlenecks in performance, energy consumption, and latency, ultimately impacting the scalability and real-world applicability of AI systems in various domains.