On Tuesday, 15/12, Dr. Dimitrios Giataganas, Department of Physics, National and Kapodistrian University of Athens, will give a seminar on “Stochastic Neural Networks as Thermodynamic Physical Systems”. The seminar will take place in the Seminar room (“Themis Paradelis”) of INPP (Tandem Building) at 14:00. [pdf]
Abstract: Machine learning has been recently used as a very effective tool for the study and prediction of data in various fields of physics. Despite its enormous success, relatively little is understood theoretically about why these techniques are so successful. A possible starting point for a theoretical treatment suggests that deep learning is a form of coarse graining. On the other hand, in theoretical physics the notion of coarse graining is characterized by the renormalization group flow. We discuss the exciting possibility of an underlying fundamental relation between certain machine learning methods and the renormalization group flow in theoretical physics. To do so we will be introducing the restricted Boltzmann machines and their training methods on lattice spin models. While the stochastic neural networks have no direct knowledge about the Hamiltonian and the interactions of the physical models, we will show that they identify spontaneously the phase transitions of the lattice spin models with a process that turns out to resemble the renormalization group flow.
Indico Page: https://indico.cern.ch/event/975635/
(Details on videoconference link will be available in the indico page)
Info : Dr. Minos Axenides, phone: +30 210 6503440 , e-mail: axenides AT inp DOT demokritos DOT gr