[ibayesclub.beta] International Bayes Club, next Thursday

[Andrade Serra, P.J. de (Paulo Jorge)]

Dear subscribers,


Next week, on Thursday the 21st, 15:00-17:00, there is a meeting of the International Bayes Club. The speakers are Pierre Alquier (RIKEN AIP), and Guang Cheng (UCLA). You can find the respective titles and abstracts below.

You can join the meeting via zoom: 
https://vu-live.zoom.us/j/93507541253?pwd=VWtibnVSbHc2Y1B4cWl2T2dLajVTZz09 
(Meeting ID: 935 0754 1253, Passcode: 299396)

As always, up to date information can be found at the Bayes Club website:
https://www.math.vu.nl/thebayesclub/


Best wishes,
Paulo

PS: sorry for any cross-posting.


---

Pierre Alquier (RIKEN AIP, JP)
Title: PAC-Bayes bounds and contraction of the posterior
Abstract:
PAC-Bayes bounds were developed to understand the generalization ability of randomized predictors, ensemble methods, and Bayesian machine learning algorithms. On the other hand, the asymptotic theory of Bayesian estimation developed tools to derive rates of contraction of the posterior in high-dimensional models. At first sight, even though they study related families of algorithms, these theories seem to have different objectives and thus rely on different mathematical tools. However, some of their assumptions are also quite similar. In this talk, I will highlight some similarities and differences between the two approaches. They can derive mutual benefit from each other. For example, variational approximations might be easier to think of in terms of PAC-Bayes bounds first, it is then possible to prove contraction results.

---

Guang Cheng (UCLA, USA)
Title: A Statistical Journey through Trustworthy AI
Abstract:
Our lab believes that the next generation of AI is mainly driven by trustworthiness, beyond performance. This talk attempts to offer statistical solutions to embrace three challenges in trustworthy AI: privacy, robustness and fairness. Specifically, we consider privacy protection by machine un-learning, enhanced adversarial robustness by utilizing unlabelled data, and establishing fair Bayes-optimal classifiers. These results demonstrate the unique value of statisticians in studying trustworthy AI from empirical, methodological or theoretical aspects. Part of this talk are based on the following works: https://arxiv.org/pdf/2202.06996.pdf, http://proceedings.mlr.press/v130/li21a/li21a.pdf, https://arxiv.org/pdf/2202.09724.pdf.

---