Gradient flow of the stochastic relaxation on a generic exponential family
Abstract
We study the natural gradient flow of the expected value E{sub p} [f] of an objective function f for p in an exponential family. We parameterize the exponential family with the expectation parameters and we show that the dynamical system associated to the natural gradient flow can be extended outside the marginal polytope.
 Authors:
 Collegio Carlo Alberto, Via Real Collegio 30, 10024 Moncalieri, Italy and Shinshu University, 4171 Wakasato, Nagano 3808553 (Japan)
 Collegio Carlo Alberto, Via Real Collegio 30, 10024 Moncalieri (Italy)
 Publication Date:
 OSTI Identifier:
 22390868
 Resource Type:
 Journal Article
 Resource Relation:
 Journal Name: AIP Conference Proceedings; Journal Volume: 1641; Journal Issue: 1; Conference: MAXENT 2014: Conference on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, Clos Luce, Amboise (France), 2126 Sep 2014; Other Information: (c) 2015 AIP Publishing LLC; Country of input: International Atomic Energy Agency (IAEA)
 Country of Publication:
 United States
 Language:
 English
 Subject:
 71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSICS; EXPECTATION VALUE; FUNCTIONS; PARAMETRIC ANALYSIS; RELAXATION; STOCHASTIC PROCESSES
Citation Formats
Malagò, Luigi, Email: malago@shinshuu.ac.jp, and Pistone, Giovanni, Email: giovanni.pistone@carloalberto.org. Gradient flow of the stochastic relaxation on a generic exponential family. United States: N. p., 2015.
Web. doi:10.1063/1.4905998.
Malagò, Luigi, Email: malago@shinshuu.ac.jp, & Pistone, Giovanni, Email: giovanni.pistone@carloalberto.org. Gradient flow of the stochastic relaxation on a generic exponential family. United States. doi:10.1063/1.4905998.
Malagò, Luigi, Email: malago@shinshuu.ac.jp, and Pistone, Giovanni, Email: giovanni.pistone@carloalberto.org. 2015.
"Gradient flow of the stochastic relaxation on a generic exponential family". United States.
doi:10.1063/1.4905998.
@article{osti_22390868,
title = {Gradient flow of the stochastic relaxation on a generic exponential family},
author = {Malagò, Luigi, Email: malago@shinshuu.ac.jp and Pistone, Giovanni, Email: giovanni.pistone@carloalberto.org},
abstractNote = {We study the natural gradient flow of the expected value E{sub p} [f] of an objective function f for p in an exponential family. We parameterize the exponential family with the expectation parameters and we show that the dynamical system associated to the natural gradient flow can be extended outside the marginal polytope.},
doi = {10.1063/1.4905998},
journal = {AIP Conference Proceedings},
number = 1,
volume = 1641,
place = {United States},
year = 2015,
month = 1
}
DOI: 10.1063/1.4905998
Other availability
Save to My Library
You must Sign In or Create an Account in order to save documents to your library.

Large deviations in stochastic heatconduction processes provide a gradientflow structure for heat conduction
We consider three onedimensional continuoustime Markov processes on a lattice, each of which models the conduction of heat: the family of Brownian Energy Processes with parameter m (BEP(m)), a Generalized Brownian Energy Process, and the KipnisMarchioroPresutti (KMP) process. The hydrodynamic limit of each of these three processes is a parabolic equation, the linear heat equation in the case of the BEP(m) and the KMP, and a nonlinear heat equation for the Generalized Brownian Energy Process with parameter a (GBEP(a)). We prove the hydrodynamic limit rigorously for the BEP(m), and give a formal derivation for the GBEP(a). We then formally derivemore » 
On Exponential Hedging and Related Quadratic Backward Stochastic Differential Equations
The dual optimization problem for the exponential hedging problem is addressed with a cone constraint. Without boundedness conditions on the terminal payoff and the drift of the Itotype controlled process, the backward stochastic differential equation, which has a quadratic growth term in the drift, is derived as a necessary and sufficient condition for optimality via a variational method and dynamic programming. Further, solvable situations are given, in which the value and the optimizer are expressed in closed forms with the help of the ClarkHaussmannOcone formula.