[theory-seminar] Theory/ML Seminar (TOMORROW 10/16): Elad Hazan

[Ofir Geri]

Hi all,


Tomorrow (Tuesday 10/16) we will have a joint theory/machine learning seminar, where Elad Hazan (Princeton) will be speaking on Taking Control by Convex Optimization (abstract below). The talk will be at 3:00 PM in Gates 463A. There will be no seminar on Friday.


Hope to see you there!

Ofir


Taking Control by Convex Optimization
Speaker: Elad Hazan (Princeton)

Linear dynamical systems, a.k.a. Kalman filtering, are a class of time-series models widely used in robotics, finance, engineering, and meteorology. In it's general form (unknown system), learning LDS is a classic non-convex problem, typically tackled with heuristics like gradient descent ("backpropagation through time") or the EM algorithm.

I will present our new "spectral filtering" approach to the identification and control of discrete-time general linear dynamical systems with multi-dimensional inputs, outputs, and a latent state. This approach yields a simple and efficient algorithms for low-regret prediction (i.e. asymptotically vanishing MSE) as well as finite-time control.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.stanford.edu/pipermail/theory-seminar/attachments/20181015/abcfb3ad/attachment.html>