[Seminar] "Relation of machine learning and renormalization group in Ising model"

Relation of machine learning and renormalization group in Ising model

 

Abstract:

Recently the machine learning has been rapidly developed, and attracts attention from researchers in various fields. The machine learning can be roughly classified into two categories: In “supervised learning” we give a machine answers for input data, while in “unsupervised learning” we don't give answers but a machine extracts some features of input data.

 

Especially, in unsupervised learning of picture images, some researchers suggest that similar phenomena to the coarse graining happen, and they point out its relation to the renormalization in particle physics.

 

In this talk, we prepare a lot of black-and-white images of the spin configurations using 2d Ising model, make a machine learn them by  unsupervised learning, and then discuss a relation with renormalization group. As a result, we find phenomena apparently different from 

renormalization group.