Student Seminar – Andres Stump and Jitze Hoogeveen on “Maximum likelihood fits with general weight matrices” and “From Radial Action to Classical Observables”

Speaker
Andres Stump and Jitze Hoogeveen

Date/Time
Date(s) - 18/12/2024
11:00 - 12:00

Location
IRIS Building, Room 1.221


Speaker: Andres Stump

Title: Maximum likelihood fits with general weight matrices

Abstract:
To perform a correlated maximum likelihood fit, one needs to determine the covariance matrix from the data. But noisy data can lead to a very ill-conditioned and sometimes non positive definite estimate for this matrix. Therefore, a weight matrix that is different from the one obtained from the covariance matrix is often used, e.g. through an SVD cut. In that case, it is not clear how goodness-of-fit parameters such as the reduced χ² and the p-value can be obtained. In this talk, I will derive formulas for these two quantities for fits with a general weight matrix.

 

Speaker: Jitze Hoogeveen

Title: From Radial Action to Classical Observables

Abstract:
Obtaining dynamics of a classical system boils down to finding the minima of the action, in phase-space. For a generic system, this action can depend on all phase-space variables, and finding solutions is not practical.
 However for special, separable systems the story simplifies dramatically. 
Recently conjectured by Riccardo Gonzo and Canxin Shi in [2405.09687], the full perturbative dynamics only depends on the radial part of the action. 
In this talk, we review this method. We discuss the underlying framework, why the radial action encodes dynamics, and how it may be explicitly evaluated.
Throughout, we treat gravitational dynamics in Kerr as an example, however conjecturally the framework applies to generic separable systems.