Revisiting natural gradient for deep networks Yoshua Bengio Universit´e de Montr´eal Montr´eal QC H3C 3J7 Canada

First Page		Document Content
Date: 2014-02-18 01:51:44 Estimation theory Numerical linear algebra M-estimators Stochastic optimization Least squares Conjugate gradient method Nonlinear conjugate gradient method Gradient descent Gradient method Hessian matrix Fisher information BroydenFletcherGoldfarbShanno algorithm		Revisiting natural gradient for deep networks Yoshua Bengio Universit´e de Montr´eal Montr´eal QC H3C 3J7 Canada Add to Reading List Source URL: arxiv.org Download Document from Source Website File Size: 669,56 KB Share Document on Facebook

	Singular Perturbations of Systems Controlled by Energy-Speed-Gradient Method DocID: 1urGY - View Document
	A Stochastic Gradient Method with an Exponential Convergence Rate for Finite Training Sets Nicolas Le Roux SIERRA Project-Team INRIA - ENS DocID: 1u6wI - View Document
	Adobe Technical Report, MarchA Perceptually Motivated Method to Control Reconstruction Errors in Gradient-based Image Compositing Sylvain Paris Abstract DocID: 1tEfO - View Document
	Steepest Descent Method Kefu Liu Properties of Gradient Vector The gradient vector of a scalar function f ( x1 , x2 ,", xn ) is defined as a column vector ⎡ ∂f DocID: 1tbCd - View Document
	SAGA: A Fast Incremental Gradient Method With Support for Non-Strongly Convex Composite Objectives Francis Bach INRIA - Sierra Project-Team DocID: 1sTsV - View Document