BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Sabre//Sabre VObject 4.5.8//EN CALSCALE:GREGORIAN BEGIN:VTIMEZONE TZID:Europe/Zurich X-LIC-LOCATION:Europe/Zurich TZURL:http://tzurl.org/zoneinfo/Europe/Zurich BEGIN:DAYLIGHT TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST DTSTART:19810329T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET DTSTART:19961027T030000 RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT UID:news1779@dmi.unibas.ch DTSTAMP;TZID=Europe/Zurich:20250219T111030 DTSTART;TZID=Europe/Zurich:20250227T161500 SUMMARY:Perlen-Colloquium: Prof. Mathias Drton (Technical University Munich ) DESCRIPTION:The ultimate aim of many data analyses is to infer cause-effect relationships between random variables of interest. While much of the ava ilable methodology for addressing causal questions relies on structural ca usal models\, these models are best suited for systems without feedback lo ops. Extensions to accommodate feedback have been proposed but often resul t in models that are challenging to interpret.\\r\\n\\r\\nIn this talk\, w e present an alternative approach to structural causal modeling: graphical continuous Lyapunov models. This framework offers a novel perspective on modeling causally interpretable dependence structures in multivariate data by treating each independent observation as a one-time cross-sectional sn apshot of an underlying temporal process. Specifically\, we focus on model s based on multivariate Ornstein-Uhlenbeck processes in equilibrium\, whic h yield Gaussian distributions where the continuous Lyapunov equation dete rmines the covariance matrix. Within this framework\, each graphical model assumes a sparse drift matrix whose support is encoded by a directed grap h.\\r\\nWe will discuss initial results on the identifiability of sparse d rift matrices and explore methods for their regularized estimation\, highl ighting the potential as well as challenges in developing graphical contin uous Lyapunov models as an interpretable tool for causal analysis in syste ms with feedback loops. X-ALT-DESC:

The ultimate aim of many data analyses is to infer cause-effe ct relationships between random variables of interest. While much of the a vailable methodology for addressing causal questions relies on structural causal models\, these models are best suited for systems without feedback loops. Extensions to accommodate feedback have been proposed but often res ult in models that are challenging to interpret.

\n\n

In this talk\, we present an alternative approach to structural causal modeling: graphica l continuous Lyapunov models. This framework offers a novel perspective on modeling causally interpretable dependence structures in multivariate dat a by treating each independent observation as a one-time cross-sectional s napshot of an underlying temporal process. Specifically\, we focus on mode ls based on multivariate Ornstein-Uhlenbeck processes in equilibrium\, whi ch yield Gaussian distributions where the continuous Lyapunov equation det ermines the covariance matrix. Within this framework\, each graphical mode l assumes a sparse drift matrix whose support is encoded by a directed gra ph.

\n

We will discuss initial results on the identifiability of spar se drift matrices and explore methods for their regularized estimation\, h ighlighting the potential as well as challenges in developing graphical co ntinuous Lyapunov models as an interpretable tool for causal analysis in s ystems with feedback loops.

END:VEVENT END:VCALENDAR