Pozvánka na přednášku semináře

  • Jan Kalina (Institute of Computer Science, Prague):

    Non-standard situations in regression modeling: From method selection to neural networks

    05.11.2024 14:00Místnost 318 (zoom) @ Ústav informatiky
    Pod Vodárenskou věží 2
    Praha, 182 00
    Hora Informaticae

    This presentation explores non-standard situations in regression modeling, emphasizing the importance of selecting appropriate methods for data analysis. Using a world tourism dataset, we demonstrate how linear regression with the commonly used Aitken estimator fails in the presence of heteroscedasticity, while robust regression quantiles offer a more suitable solution. In the second part, we examine robust versions of regularized neural networks, explaining how their regularization can be interpreted through the framework of Bayesian inference under specific assumptions. This Bayesian perspective offers a likelihood-based approach to training neural networks with prior information.

  • Štěpán Holub (Charles University, Faculty of Mathematics and Physics):

    Isabelle/HOL and Its Logic: User's Point of View

    06.11.2024 14:00Místnost ZOOM, Institute of Philosophy @
    Seminář aplikované matematické logiky

    Isabelle is a generic proof assistant that allows axiomatization of arbitrary object logic, the most widely used of which is HOL. The talk will be an introduction in Isabelle/HOL. I will briefly sketch its history and mention my motivation for using it in an ongoing project of formalization of combinatorics on words. The main part of the talk will be a live session illustrating how some design features of Isabelle/HOL are reflected in user's work. Time permitting, these will include the distinction between meta-logic and the object logic, polymorphic simple types, higher order unification, (non-existence of) proof terms, design of proof tactics, and construction of verified algorithms.

Předchozí přednášky