Please save the date and welcome FRF Candidate, Sarah Harvey, Stanford University who will be presenting on March 10th at 10:30AM in the 4th Floor Classroom at 160 5th Avenue.
Please contact me if you wish to meet with Sarah 1:1 between March 10th – 12th as she will be visiting with us at CCN for these days.
Title: Combating Noise and Uncertainty in Biophysical Models
Abstract: Noise and uncertainty are ubiquitous in biological systems, and robustness to these effects may be a crucial piece of understanding biological design. In small systems, thermal fluctuations can be of the order of the energy difference between system states, and these fluctuations are an important operational consideration for systems at the mesoscale. At larger scales, biological systems like neural networks are assembled using large numbers of seemingly noisy components. At the behavioral level, organisms constantly confront an unpredictable world and must make decisions that achieve their goals but are also sensitive to risk. I will discuss three projects that attempt to provide insight into robust biophysical models at each of these scales. At the smallest scale, we investigate theoretical bounds on the accuracy of single cellular sensors and how this is limited by energy dissipation. At the level of neural networks, we propose a local learning rule that provably increases the network robustness to noise while preserving task performance. Lastly, we apply large deviation theory to risk-sensitive reinforcement learning in order to generate variance-constrained policies.
Zoom information in the Google Calendar Invite.