Byron Yu

News:

Research interests:

1. Statistical techniques for studying neural mechanisms in large-scale neural recordings
   
   Advances in neural recording technology have enabled unprecedented views of the simultaneous activity of neural populations in vivo. How can we obtain scientifically meaningful depictions of neural processes from these noisy, high-dimensional data? What does this tell us about the computations performed by the underlying neural circuitry?


2. Neural basis of motor preparation and execution
   
   How do populations of neurons guide natural arm movements? What aspects of these movements be prepared or "planned" well in advance of movement execution? How are movement plans then converted into physical movements? We combine electrophysiological (both multi-electrode arrays and single electrodes), behavioral, and computational approaches to study such questions.


3. Neural prosthetic system design
   
   Neural prosthetic systems aim to assist paralyzed patients by translating their thoughts into limb movements. Given a population of neurons, what is the "best" neural decoder that can be designed, in terms of decoding accuracy and speed, computational complexity, robustness to neuron loss, and suitability for the desired usage mode(s)?


4. Machine learning approaches in neuroscience
   
   I am particularly interested in latent variable methods applied to neural data. I have previously worked on approximate inference and learning algorithms for nonlinear dynamical systems, including those based on expectation-propagation (EP), quadrature integration, and Laplace's method.

Links: