# MSCS Seminar Calendar

Wednesday May 6, 2015

**Computer Science Seminar**

Convex Risk Minimization and Conditional Probability Estimation

Matus Telgarsky (University of Michigan)

3:00 PM in SEO 427

This talk will investigate and strengthen the link between convex risk
minimization and conditional probability estimation, a connection
already notable for establishing consistency results (Friedman et al.,
2000; Zhang, 2004; Bartlett et al., 2006). Specifically, this talk will
first show that a differentiable loss function, linear space of
predictors, and probability measure together define a unique optimal
conditional probability model, moreover one which may be attained by the
usual convex risk minimization. This result is proved in infinite
dimensions, and thus gives a concrete convergence target for
unregularized methods like boosting which can fail to have minimizers.
Second, this convergence result is refined in finitely many dimensions
to hold for empirical risk minimization. This uniform convergence
result exhibits no dependence on the norms of its predictors, and thus
can justify the practical effectiveness of minimally-regularized
optimization schemes.
This is joint work with Miroslav Dudik and Robert Schapire

Thursday May 21, 2015

**Algebraic Topology Seminar**

Takayasu's cofiber sequences of Thom spectra over the classifying space of ${(\mathbb{Z}/2)}^n$

Maximilien Holmberg-Péroux (EPFL/UIC)

11:00 AM in SEO 1227

After presenting virtual vector bundles and their associated Thom spectra, I will present a modern point of view of the cofibre sequence of Takayasu’s paper « On stable summands of Thom spectra of $B(\mathbb{Z}/2)^n$ », its motivational aspect, and a sketch of the proof.

Saturday May 23, 2015

Friday August 28, 2015