ELFT HÍRADÓ

[StatFizSzeminar <statfiz AT glu.elte.hu>]

            #    #    #    #    #    #


             ELTE TTK Fizikai Intézet
         STATISZTIKUS FIZIKAI SZEMINÁRIUM


                  2018. október 3.
                   szerda, 11.00 

                 Michael C. Abbott              

              Jagiellonian University

"Maximizing the information learned from finite data 
              selects a simple model"

We use the language of uninformative Bayesian prior
choice to study the selection of appropriately simple
effective models. We advocate for the prior which max-
imizes the mutual information between parameters and 
predictions, learning as much as possible from limited
data. When many parameters are poorly constrained by 
the available data, we find that this prior puts weight
only on boundaries of the parameter manifold. Thus it
selects a lower-dimensional effective theory in a prin-
cipled way, ignoring irrelevant parameter directions. 
In the limit where there is sufficient data to tightly
constrain any number of parameters, this reduces to 
Jeffreys prior. But we argue that this limit is patho-
logical when applied to the hyper-ribbon parameter man-
ifolds generic in science, because it leads to dramatic
dependence on effects invisible to experiment.

1117, Budapest, Pázmány P. sétány 1/A, Északi tömb 2.54
        honlap: http://glu.elte.hu/~statfiz

            #    #    #    #    #    #