Research Themes

Model complexity is a key design challenge today, especially in the context of deployment to mobile and ultra-low-power hardware platforms. We propose to pursue a wide variety of projects with the goal of improving machine learning training and inference efficiency which are categorized to two main themes as follows:

• Hardware-Software Co-design of Heterogenous Neural Network Systems (PIs: Prof. Warren Gross and Prof. Brett Meyer)
• Design of Efficient Well-Calibrated and Adaptable Deep Neural Networks (PI: Prof. James Clark)