Computational Neuroscience

2 Credits · Level: Graduate, Undergraduate

Seemingly without effort, the human brain solves computationally very complex cognitive tasks such as recognizing facial expressions, understanding speech, planning and executing sequences of movements, or choosing actions from a range of competing alternatives to maximize the likelihood of reward in a given situation. How does the brain do it? For instance, how can signals from one hundred million photoreceptors on the retina fit through an optic nerve with just one million fibers? Given that many cognitive tasks are too varied to be coded genetically, how are tasks learned in the “deep” processing hierarchy of the brain, with some 13-odd synapses in the decision pathway? How does the brain learn to associate actions and rewards? How can we make decisions in the face of uncertainty? How does the brain deal with the redundancy of the motor plant, coordinating many joints to reach points in three-dimensional space? How can the brain recall complete memories from partial cues? This class will examine these and related questions, showing how cognitive processing across a variety of areas can be grounded in a small number of key neurocomputational principles. Emphasis will be placed on showing how these principles apply to a variety of cognitive domains, including vision, audition, memory, motor control, and decision making. The goal of this introductory class is to convey the underlying computational ideas with a minimum of mathematical overhead, stressing their usefulness even for areas of cognition where data are still insufficient to constrain quantitative models. Course Description Credits: 2 Time: Tuesdays, 1:15pm – 4pm Dates: Class starts on 9/1/2020, until 12/1/2020 Location: Neuroscience Meeting Room, Research Building, WP-07 Course Director: Maximilian Riesenhuber, PhD, Professor of Neuroscience (mr287@georgetown.edu) Office: New Research Building, WP-12 Open to graduate and advanced undergraduate students from neuroscience, psychology, cognitive science, computer science, and related disciplines. Undergraduate students should contact the course director for permission. Prerequisites Basic knowledge of neuroscience and/or psychology. Basic calculus and linear algebra. Willingness to learn whatever background is lacking. Contact the course director if unsure. Course Structure and Requirements This is a 2-credit course, but classes meet for three hours each, which allows for off-weeks when we won't meet, see the schedule below and the course schedule on the Canvas site. I highly recommend you add the Canvas online calendar to your personal calendar. Each class will consist of a lecture by Dr. Riesenhuber on the topic du jour and student-led presentations and discussions of research papers (4 papers of 20-30 minutes of presentation & discussion each), see “Schedule and Papers” below (note that the “Optional background papers” are optional further readings for those interested in particular topics). These presentations should include critical analyses of the methods and results, along with a discussion of the implications of the findings and three questions to be discussed with the class. Presenters should feel free to discuss questions about their papers with the course director in advance of their presentation. All class members will be expected to have read these papers before class and to come prepared with relevant questions/comments/ideas for the discussion. In addition, there will be (optional) computer exercises on key topics in the form of Jupyter notebooks that students can work on outside of class. A final project will be due on December 1. The final project will take the form of a 15-minute presentation on an original computational neuroscience project of the student’s choosing. The project should aim to test (either experimentally or through simulations or a combination thereof) a computational hypothesis about the neural bases of a particular cognitive function. The presentation (to be emailed to Dr. Riesenhuber at mr287@georgetown.edu by midnight on December 1 – each day late takes off 5% of final project score) needs to include specific aims of the research project, background information about the topic studied, and a description of a series of experiments/computational studies to address the Specific Aims. The presentation must include a discussion of what conclusions could be drawn depending on various outcomes of the experiments. An outline of the project (1 page of Specific Aims, due November 17, but earlier submission is encouraged if students already have a project idea) must be approved by Dr. Riesenhuber so that students can receive feedback on the appropriateness of their chosen topic before working on the presentation. Participants’ grades will depend on 4 paper presentations (40%), participation in class discussions (20%), and the final project (40%).