A research proposal by Anahita Khojandi, assistant professor in ISE, earned funding this month from the Science Alliance’s Joint Directed Research and Development (JDRD) Program. Her project, titled “Dynamic Deep Reinforcement Learning-Bayesian Framework,” will receive a budget of $125 thousand through July 31, 2020.
The project is aligned with Oak Ridge National Laboratory’s current AI Initiative.
“The goal of the proposed work is to develop a holistic framework for decision making using multi-channel, high-frequency data streams,” said Khojandi.
This real-time framework will integrate deep-reinforcement learning (DRL) and Bayesian modeling, while accounting for various human-in-the-loop considerations and resource limitations. It will also advance the knowledge on developing and improving DRL algorithms that can be exploited to efficiently solve the proposed framework.
The research will benefit a variety of applications, notably in healthcare.
“For instance, the framework can leverage high-frequency health data, collected from bedside monitors, to improve early sepsis detection performance,” said Khojandi. “Or it can be adopted to take advantage of wearable sensor data collected from Parkinson’s disease patients to optimize their individualized treatment plans, among others.”
The JDRD program is designed to focus on strategic research interests while enhancing collaborations between UT faculty and ORNL scientists to address problems with far-reaching impacts.