Professor Hadi Esmaeilzadeh, the Halicioğlu Chair in Computer Architecture, and Ph.D. students Byung Hoon Ahn and Sean Kinzer were nominated for a Design Automation Conference (DAC) Best Paper Award. Their research, Glimpse: Mathematical Embedding of Hardware Specification for Neural Compilation, was chosen as one of six nominees for this prestigious award.
The CSE research team’s paper proposes a previously unexplored approach to optimizing the code for Deep Neural Network (DNN) hardware. They devised a Bayesian optimization framework called Glimpse with multiple exclusively unique components, allowing them to mathematically embed the hardware specifications into the search. When Glimpse was compared with hardware-agnostic compilers, the results indicated faster compilation time and the best inference latency