Data Analysis Project Presentation

  • Ph.D. Student
  • Joint Ph.D. Program in Neural Computation and Machine Learning
  • Carnegie Mellon University
Project Presentations

Inferring attentional state using behavior and neural activity

Subjects in a covert spatial attention experiment are typically directed to pay attention to one of two locations in space. Typically, studies make the simplifying assumption that the subject always follows the instruction of the experimenter, and that the level of attention is constant from trial to trial. However, attentional state likely varies greatly from trial to trial, and is sometimes allocated to the incorrect location. Recent work has focused on using an "attention axis" in neural activity space to infer a graded attentional state on a single trial basis (Cohen & Maunsell, 2010). Here, we extend this framework by incorporating behavioral variables and task parameters into a probabilistic graphic model to infer a latent attentional state. We then use the inferred latent to create a modified attention axis with better out-of-sample prediction of behavior.

Faculty Committee:
Max G'sell, Byron Yu, Matthew Smith

For More Information, Please Contact: