This is an online seminar. Registration is required.
We’ll send the instruction for attending the online seminar.
【Deep Learning Theory Team】
【Date】2021/Feb/16 (Tue)
Title: Some results on stochastic gradient type algorithms
Abstract: In the first part, we discuss some results for the sampling problem using Langevin dynamics, in particular, the Stochastic Gradient Langevin Dynamics (SGLD) algorithm. We allow the estimation of gradients to be performed even in the presence of dependent data streams.
Non-asymptotic analysis are established in appropriate Wasserstein distances for the convex and non-convex cases. Our convergence estimates are sharper and uniform in the number of iterations, in contrast to those in previous studies. In the second part, other achievements are discussed.
https://zoom.us/j/96515536515?pwd=UmFmUWhjOGpGdEFlMmpIbEMrOU05dz09
ミーティングID: 965 1553 6515
パスコード: 033235
