Please join us for our virtual seminar on Thursday, December 16th with guest speaker, Sarah Marzen, Assistant Professor of Physics, W. M. Keck Science Department, Pitzer, Scripps, and Claremont McKenna College.
Title: "Everything's A Reservoir: Using the Back of the Brain to Improve Machine Learning Prediction Algorithms"
Abstract: The brain is thought to predict input efficiently and well. We add to the evidence that the brain memorizes in order to predict, focusing on time cells in the hippocampus. We then use known response properties of predictors in the brain to improve machine learning algorithms, increasing the prediction function of nonlinear reservoirs trained on natural video. This points the way to future work that not only reinterprets parts of the brain as memorizers for the purpose of prediction, but also uses properties of the brain to improve more complicated machine learning prediction algorithms.
Please feel free to convene in the 9th Floor Classroom if you are onsite, and if tuning in remotely, please use the zoom credentials listed on the calendar invite.