Page 9 - Demo
P. 9
Scale lengths are obtained from data using Occam’s razor
Evidence maximiza@on respect to scale lengths1
If forward model is linear D= K I :
L = log p(D I,θ)= −1logKΣKT +Σ −1DT(KΣKT +Σ )D−Nlog(2π)
Simula@on data (LR Equilibrium Code 2) vs inference results
ID 2ID22
!###"###$ !###"####$ !#"#$
Model Complexity Data Fit
Offset
1C.E.Rasmussen , C. Williams ( 2006), Gaussian Processes for machine learning. The MIT Press 2L. Galeo\\ et al (2011) Physics of Plasmas 18, 082509.
Presenter : J.Romero. Sesion T07. Computational and theoretical Techniques. APS 2016.
9
