Biomedical Engineering

Statistics of Memory: Towards Closed Loop Interaction With the Hippocampus

Printer-friendly versionSend by emailPDF version


Caleb Kemere, Ph.D.


Friday, March 24, 2017 - 12:00pm


SEC 204


One of the more remarkable aspects of neural circuits in the brain is that they are flexible - depending on underlying state changes like arousal and attention, neural ensembles may represent information in very different ways. Systems which facilitate state-dependent manipulations of these neural circuits will enable us to understand how the different modes of operation of a circuit support the learning and execution of behaviors that are critical for an organisms daily life. I will talk about two recent advances we've made as we've built and analyzed real-time systems which interact with the neural circuits of memory. We are interested in detecting and decoding patterns associated with memory reactivation that occur during bursts of activity in the hippocampus called sharp wave ripples. I'll present a recent full system analysis that has helped us optimize or real time system as well as making its limits clear, and serves as a case study for such experiments. In addition, I'll show how a computationally efficient latent variable model allows us to decode neural activity as well as understand the noise statistics of the information represented in the neural circuit.