Title: Machine Learning for Large-scale Neuroscience and Mental Health: Opportunities and Challenges
Abstract: In recent years, machine learning (especially deep learning) has been successfully applied in many domains, including computer vision, speech recognition, gaming, autonomous driving, document analysis etc. To date, advances in neurotechnologies have allowed us to record large-scale, high-throughput neural recordings (e.g., electrophysiology, calcium imaging) and animal behaviors, which also present opportunities and challenges in data analysis and real-time applications, such as closed-loop brain-machine interfaces (BMIs). Here, we discuss how to use machine learning to address challenging neuroscience (BIG data) and medicine (SMALL data) problems, including neural decoding, signal detection, and psychiatric diagnosis. Specifically, we illustrate a few principles using several real-world examples carried out in my laboratory.
(i) Machine learning for real-time rodent neural decoding based on calcium imaging (~1000 cells) or local field potentials (~512 channels) from the rodent hippocampus. (ii) Deep learning for human/rodent EEG sleep-spindle detection and its application in closed-loop BMI application. (ii) Deep learning for human psychiatric diagnosis based on fMRI recordings. Finally, we discuss research challenges related to several important issues on transfer learning, continual learning, label imbalance and error. We also discuss future opportunities in neuroscience (multimodality) and mental health (mobile e-health).
|日時||2020/09/30(水) 14:00 - 15:00|