Final Report on Accelerator Architectures and Algorithms for Embedded Machine Learning in Medical Sensor Applications
Research Report Highlight
Researchers from Princeton developed six medical sensor applications using a configurable, machine learning accelerator. A 3D, face to face IC architecture was employed to enable high accelerator/memory bandwidth and a 3 to 500 x energy reduction.
This report provides testing details and characterization summary of a 3D IC for accelerator-based architectures. This report also summarizes the major research accomplishments and future directions from the overall project.