Currently I have two major interests. First, adversarial attack and defense of deep learning models. Specifically we are trying to craft adversarial examples in the physical world sensed by different equipments including visible light cameras and infrared cameras. In this way, we expose the vulnerability and risk of deep learning models. We're also looking for ways to get more robust deep learning models.
Second, brain-inspired computational models. We develop new models to circumvent the difficulties that current deep learning models are facing including the above stated vlnerability. Towards this goal, we are collaborating with neuroscientists to unravel the neural wiring diagram of the mouse brain.
Besides the above main interests, I'm also interested in AI+robotics.