| Document Date: 2013-08-04 20:52:03
Open Document File Size: 464,63 KBShare Result on Facebook
City Atlanta / / Company Neural Information Processing Systems / NA HF / Neural Networks / Recurrent Neural Networks Echo-State Networks / MIT Press / / Country United States / Spain / / Currency SGD / / / Facility University of Toronto / / IndustryTerm mini-batch algorithms / faster stochastic gradient search / deep belief networks / deep and recurrent network / iterative algorithm / recurrent networks / linear conjugate gradient algorithm / online learning / default initial solution / temporal networks / initialized networks / deeper networks / computing / feedforward networks / information processing systems / chaotic systems / stochastic search / wireless communication / tanh networks / learning algorithm / energy / using deep belief networks / / Organization MIT / University of Toronto / / Person Le / George Dahl Geoffrey Hinton / Kombrink / James Martens George Dahl Geoffrey / / Position author / General / / ProvinceOrState Nevada / Georgia / / PublishedMedium Journal of Physics A / Machine Learning / / Technology LAN / speech recognition / neural network / linear conjugate gradient algorithm / mini-batch algorithms / http / iterative algorithm / Machine Learning / two algorithms / /
SocialTag |
On the importance of initialization and momentum in deep learning Ilya Sutskever1 James Martens George Dahl Geoffrey Hinton