[Ml-stat-talks] Course of interest
powell at princeton.edu
Tue Sep 12 19:36:13 EDT 2017
To the machine learning community...
If you are looking for a new machine learning challenge, you might consider
sitting in on the first lecture of ORF 544 - Stochastic optimization which
will meet tomorrow (Wednesday) at 1:30pm in Friend 004.
Stochastic optimization is *much* richer than reinforcement learning (that
you may have heard of). Another special class of problems are the
multiarmed bandit problems.
These are just two classes (out of about 15) of stochastic optimization
problems, which I will model using a single unified framework.
I will show in the course that you can solve *any* stochastic optimization
problem using one of four classes of policies. Three of these involve
machine learning, and there are four types of functions that need to be
approximated using machine learning.
However, there are a number of twists that make these machine learning
different than the classical "big data" problems. In fact, there is a lot
of online learning.
The course will focus on careful modeling and practical algorithms rather
than the kinds of theory (convergence proofs, bounds) that are popular in
these fields. Listeners are welcome, but even if you do not have time to
sit in on a course, consider attending just the first lecture.
*------------------------------Warren B. Powell*
*Professor, Department of Operations Research and Financial Engineering*
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Ml-stat-talks