News & Events

Zhu’s research paper accepted by AAAI Conference on Artificial Intelligence 2021

Our new Assistant Professor Dr. Chunjiang Zhu’s research paper entitled “An Efficient Algorithm for Deep Stochastic Contextual Bandits” has been accepted to appear in the 35th AAAI Conference on Artificial Intelligence (AAAI), 2021. AAAI is a premier conference in Artificial Intelligence and Machine Learning, with about 8,000 submissions being reviewed and an acceptance ratio of 21% this year.

In the paper, Dr. Zhu and his collaborators study the problem of stochastic contextual bandits (SCB), where an agent selects an action based on certain observed context to maximize the cumulative reward over iterations. There have been several works that used a deep neural network (DNN) to predict the expected reward for an action and trained the DNN by a stochastic gradient based method. However, rigorous analysis has seldom been given to examine whether and where the DNN and the action policy converge. This paper fills the void by proposing a stage-wise stochastic gradient descent for SCB, which exploits adaptive action policies in an efficient manner and enjoys provable convergence on the action policy and reward function.

More of Dr. Zhu’s research can be found on his research webpage: