AUTHOR=Zhang Fan , Yang Li , Meng Jian , Seo Jae-sun , Cao Yu , Fan Deliang 

TITLE=XMA2: A crossbar-aware multi-task adaption framework via 2-tier masks

JOURNAL=Frontiers in Electronics

VOLUME=Volume 3 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/electronics/articles/10.3389/felec.2022.1032485

DOI=10.3389/felec.2022.1032485

ISSN=2673-5857

ABSTRACT=Recently, ReRAM crossbar-based deep neural network (DNN) accelerator has been widely investigated. However, most prior works focus on single-task inference due to the high energy consumption of weight re-programming and ReRAM cells’ low endurance issue. Adapting the ReRAM crossbar-based DNN accelerator for multiple tasks has not been fully explored. In this paper, we propose XMA2, a novel crossbar-aware learning method with a 2-tier masking technique to efficiently adapt a DNN backbone model deployed in ReRAM crossbar for new task learning. During the XMA2 based multi-task adaption, the tier-1 ReRAM crossbar-based Processing-Element (PE)-wise mask is first learned to identify the most critical PEs to be re-programmed for essential new features of the new task. Next, the tier-2 crossbar column-wise mask is applied within the rest weight-frozen PEs to learn a hardware-friendly and column-wise scaling factor for new task learning without modifying the weight values. With such crossbar-aware design innovations, we could implement the required masking operation in an existing crossbar-based convolution engine with minimal hardware/memory overhead to adapt to a new task. The extensive experimental results show that, compared with other state-of-the-art multiple-task adaption methods, XMA2 achieves the highest accuracy on all popular multi-task learning datasets.