BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//RLASKEY//CALENDEROUS//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
DTSTAMP:20260815T210410Z
LAST-MODIFIED:20170508T133144Z
DTSTART:20170510T190000Z
DTEND:20170510T200000Z
UID:event1767@bu.edu
URL:http://physics.bu.edu/internal/events/show/1767
SUMMARY:Stochastic Neural Networks for Machine Learning the Many-Body Probl
	em
DESCRIPTION:Featuring Roger Melko\n\nPart of the Condensed Matter Theory Se
	minar Series.\n\nCondensed matter physicists have a sophisticated array of 
	numerical\ntechniques that they use to study classical and quantum many-bod
	y models. \nIn parallel\, the machine learning community has developed a ve
	ry successful\nset of algorithms with the goal of classifying\, characteriz
	ing and\ninterpreting complex sets of data\, such as images and natural lan
	guage\nrecordings. We briefly show that standard neural networks architectu
	res for\nsupervised learning can identify phases and phase transitions in a
	 variety\nof condensed matter Hamiltonians\, directly from raw state config
	urations\nsampled with standard Monte Carlo and treated like images.  Then\
	, we show\nhow we can use such Monte Carlo configurations to train a stocha
	stic variant\nof a neural network\, called a Restricted Boltzmann Machine (
	RBM)\, for use in\nunsupervised learning applications.  We demonstrate how 
	RBMs\, once trained\,\ncan be sampled much like a physical Hamiltonian to p
	roduce configurations\nuseful for estimating physical observables\, as well
	 as other applications.\nFinally\, we explore the representational power of
	 RBMs\, and comment on their\napplication to the simulation of quantum syst
	ems.
LOCATION: \, \, 
STATUS:CONFIRMED
CLASS:PUBLIC
END:VEVENT
END:VCALENDAR