Research
Current Research Projects
- Continual learning & neural representations Animals and humans, due to their neuroplasticity and brains’ ability to learn and adapt, can continually learn without forgetting previously acquired knowledge. How exactly biological brains solve the problem of continual learning and catastrophic forgetting remains elusive. However, with the rise of deep learning, particularly AI, researchers and engineers are developing algorithms that learn from a continuous stream of data, acquiring new knowledge or skills over time and avoiding catastrophic forgetting. I currently explore how biological brains solve the problem of continual learning and catastrophic forgetting, combining methods from neuroscience and machine learning. I develop computational models simulating animals' behavior, analyze the neural responses and connectivity patterns of those models, and examine how they are affected or how they affect the learning process.
- Physics-Informed Neural Networks (PINNs) An emerging subfield of machine learning is scientific deep learning. This field focuses on using deep learning to solve differential equations and minimize energy functionals in physics and materials science. It also aims to discover physics itself. Building on these aims, I’m interested in developing new deep learning methods to minimize energy functionals and to solve partial, ordinary, and integrodifferential equations in neuroscience and materials science. He will leverage the tools modern deep learning offers.
Past Research Projects
-
Neuromorphic Computing I was involved in developing neuromorphic devices capable of embedded and online learning. I was one of the core developers of the software simulator for the Neural and Synaptic Array Transceiver Framework (in collaboration with Intel Corporation Research Labs, the University of California, Irvine, and San Diego). Moreover, I investigated how natural mechanisms of biological brains can lead to more efficient and biologically plausible machine learning algorithms suitable for neuromorphic devices.
-
Parkinson’s Disease Another research interest is Parkinson’s disease (PD). He has combined neuroscience and control theory to study PD and propose potential treatments. He has developed computational models based on neural field theory for the basal ganglia and PD, and has investigated closed-loop deep brain stimulation (DBS) with applications to PD treatment.
-
Cortical Plasticity & Self-organizing Maps During my Ph.D. research, I focused on cortical plasticity and self-organization. I proposed a mathematical/computational model for studying self-organization in the primary somatosensory cortex, based on neural field theory. Simulations of this model demonstrated how the cerebral cortex develops topographic maps, maintains them, and reorganizes them after a cortical lesion (e..g., stroke) or a sensory deprivation (e.g., limb amputation).
-
Rhythmical Motor Control As a Master’s student, I studied rhythmical motor control and human tremor. I applied signal processing theory and analysis to neurophysiological signals such as EEG and EMG to determine the role of human tremor. Furthermore, I investigated Central Pattern Generators (CPGs) and their role in biped locomotion on humans and humanoid robots.
He also has some experience in: