Abstract:
Brain computer interfaces (BCIs) have been widely adopted to enhance human perception via brain signals with abundant spatial-temporal dynamics, such as electroencephalogram (EEG). In recent years, BCI algorithms are moving from classical feature engineering to emerging deep neural networks (DNNs), allowing to identify the spatial-temporal dynamics with improved accuracy. However, existing BCI architectures are not leveraging such dynamics for hardware efficiency. In this work, we present uBrain, a unary computing BCI architecture for DNN models with cascaded convolutional and recurrent neural networks to achieve high task capability and hardware efficiency. uBrain co-designs the algorithm and hardware: the DNN architecture and the hardware architecture are optimized with customized unary operations and immediate signal processing after sensing, respectively. Experiments show that uBrain, with negligible accuracy loss, exhibits significant power advantage over the CPU, systolic array and stochastic computing baselines.
Bio:
Di Wu is currently a PhD candidate in Department of ECE at UW–Madison, advised by professor Joshua San Miguel. His research interest lies in the field of emerging computer architecture, including unary computing, approximate computing, reconfigurable computing and quantum computing. His research was featured in Qualcomm Innovation Fellowship Finalist in 2019 and awarded IEEE Micro Top Pick in 2021. He received bachelor and master degrees in 2012 and 2015 from Fudan University in Shanghai, China. Before starting PhD in 2017, he was a digital circuit engineer in HiSilicon. During PhD study, he did internships related to deep learning hardware software co-design in Meta and Cerebras Systems.
Read More