BEGIN:VCALENDAR
VERSION:2.0
PRODID:https://murmitoyen.com/events/vanille/udem/
X-WR-TIMEZONE:America/Montreal
BEGIN:VEVENT
UID:69dc1c3e2e66f
DTSTAMP:20260412T182710
DTSTART:20161212T113000
SEQUENCE:0
TRANSP:OPAQUE
DTEND:20161212T123000
URL:https://murmitoyen.com/events/vanille/udem/detail/717846-machine-learni
 ng-the-many-body-problem-roger-melko-u-of-waterloo
LOCATION:Université de Montréal - Pavillon J.-Armand-Bombardier\, 5155\, 
 chemin de la rampe \, Montréal\, QC\, Canada\, H3T 2B2
SUMMARY:Machine Learning the Many-Body Problem - Roger Melko (U. of Waterlo
 o)
DESCRIPTION:Roger Melko\, Department of Physics and Astronomy\, University 
 of Waterloo
\n 
\nabstract:  Condensed matter physics is the study of the
  collective behavior of infinitely-complex assemblies of electrons\, magne
 tic moments\, atoms or qubits. This complexity is reminiscent of the “c
 urse of dimensionality” commonly encountered in machine learning.  Desp
 ite this curse\, the machine learning community has developed techniques 
 with remarkable abilities to classify\, characterize and interpret complex
  sets of data\, such as images and natural language recordings. Here\, we
  show that modern architectures for supervised learning\, such as fully-co
 nnected and convolutional neural networks\, can identify phases and phase
  transitions in a variety of condensed matter Hamiltonians. Readily progr
 ammable through open-source software libraries\, neural networks can be tr
 ained to detect multiple types of order parameter\, as well as highly non
 -trivial states with no conventional order\, directly from raw state confi
 gurations sampled with standard Monte Carlo.  Further\, Monte Carlo conf
 igurations can be used to train a stochastic variant of a neural network\,
  called a Restricted Boltzmann Machine (RBM)\, for use in unsupervised le
 arning applications.  We show how RBMs\, once trained\, can be sampled mu
 ch like a physical Hamiltonian to produce configurations useful for estim
 ating physical observables. Finally\, we explore the representational powe
 r of RBMs\, their role in deep learning\, and its possible relationship t
 o the renormalization group.
\nFor more information about the research of 
 Prof. Melko\, you can consult his web page.
\n 
\nCette conférence est p
 résentée par le RQMP Versant Nord du Département de physique de l'Uni
 versité de Montréal et le Département de génie physique de Polytechn
 ique Montréal.
END:VEVENT
BEGIN:VTIMEZONE
TZID:America/Montreal
X-LIC-LOCATION:America/Montreal
END:VTIMEZONE
END:VCALENDAR