Statistics at UC Berkeley: We are a community engaged in research and education in probability and statistics. In addition to developing fundamental theory and methodology, we are actively involved in statistical problems that arise in such diverse fields as molecular biology, geophysics, astronomy, AIDS research, neurophysiology, sociology, political science, education, demography, and the U.S. Census. We have forged strong interdisciplinary links with other departments and areas of study, particularly biostatistics, mathematics, computer science, and biology, and actively seek to recruit graduate students and faculty who can help to build and maintain such links. We also offer a statistical consulting service each semester.
Statistics at UC Berkeley
Oct 26, 2020
Seminar 217, Risk Management: Improving reinforcement learning algorithms using an optimal learning rate policy (Online)
Othmane Mounjid, Ãcole Polytechnique (Speaker)
ABSTRACT: We investigate to what extent one can improve reinforcement learning algorithms. For this, we first show that the classical asymptotic convergence rate O(1/âN) is pessimistic and can be replaced by O((log(N)/N)^Beta) with Beta in [0.5,1] and N the number of iterations. Second, we propose a dynamic optimal policy for the choice of the learning rate. We decompose our policy into two...
GerÃ³nimo URIBE BRAVO, Instituto de MatemÃ¡ticas Universidad Nacional AutÃ³noma de MÃ©xico
For a given (plane) tree, let N_i be the quantity of individuals with i descendants and define its degree sequence as s=(N_i). Focus will be placed on the uniform distribution on trees whose degree sequence is s. We give conditions for the convergence of the profile (aka the sequence of generation sizes) as the size of the tree goes to infinity. This gives a more general...
Aaron Roth, University of Pennsylvania
Abstract: We show how to achieve multi-calibrated estimators not just for means, but also for variances and other higher moments. Informally, this means that we can find regression functions which, given a data point, can make point predictions not just for the expectation of its label, but for higher moments of its label distribution as well --- and those predictions match the true distribution...
Peng Ding, UC Berkeley
Fisherâs randomization test delivers exact p-values under the strong null hypothesis of no treatment effect on any units whatsoever and allows for flexible covariate adjustment to improve the power. Of interest is whether the procedure could also be valid for testing the weak null hypothesis of zero average treatment effect. Towards this end, we evaluate two general strategies for Fisher...
Zhipu Zhou, UC Santa Barbara (Speaker)