[Statlist] Research Seminar in Statistics *FRIDAY, 6 MARCH 2020* GSEM, University of Geneva

gsem-support-instituts g@em-@upport-|n@t|tut@ @end|ng |rom un|ge@ch
Mon Mar 2 09:10:29 CET 2020


Dear All,

We are pleased to invite you to our next Research Seminar.

Looking forward to seeing you


Organizers :                                                                                   
E. Cantoni - S. Engelke - D. La Vecchia - E. Ronchetti
S. Sperlich - F. Trojani - M.-P. Victoria-Feser


FRIDAY, 6 MARCH 2020 at 11:15am, Uni-Mail M 5220

Learning with Domain Knowledge
Alexandros KALOUSIS - University of Applied Sciences, Western Switzerland, and University of Geneva

ABSTRACT:
We can significantly improve the learning behavior, in terms of sample complexity and predictive performance, if within the learning process we can bring in and exploit knowledge and information in addition to the training data. In this talk I will provide an overview of the work my team is doing in order to exploit such additional knowledge to provide meaningful structure and constraints to the learning process. These can be derived from domain knowledge and can be used to impose desired structure on the learned model. We will provide examples of such settings and show how we exploit declarative constraints to regularize the learned model's structure by controlling its Jacobian. In addition to such declarative structure and constraints, we can also exploit knowledge that is encoded black-box within programming artifacts such as domain specific software packages, e.g. RDKit in chemoinformatics, and simulators. I will also review some of the work we are doing on molecule generation and molecule style transfer, using variational autoencoders, that seeks to exploit such programming artifacts.


Visit the website: https://www.unige.ch/gsem/en/research/seminars/rcs/




More information about the Statlist mailing list