Search For:

Displaying 1-10 out of 10 total
Recognizing Handwritten Digits Using Hierarchical Products of Experts
Found in: IEEE Transactions on Pattern Analysis and Machine Intelligence
By Guy Mayraz, Geoffrey E. Hinton
Issue Date:February 2002
pp. 189-197
<p><b>Abstract</b>—The product of experts learning procedure can discover a set of stochastic binary features that constitute a nonlinear generative model of handwritten images of digits. The quality of generative models learned in this w...
 
Using Generative Models for Handwritten Digit Recognition
Found in: IEEE Transactions on Pattern Analysis and Machine Intelligence
By Michael Revow, Christopher K.I. Williams, Geoffrey E. Hinton
Issue Date:June 1996
pp. 592-606
<p><b>Abstract</b>—We describe a method of recognizing handwritten digits by fitting generative models that are built from deformable B-splines with Gaussian
 
Modeling pixel means and covariances using factorized third-order boltzmann machines
Found in: Computer Vision and Pattern Recognition, IEEE Computer Society Conference on
By Marc'Aurelio Ranzato, Geoffrey E. Hinton
Issue Date:June 2010
pp. 2551-2558
Learning a generative model of natural images is a useful way of extracting features that capture interesting regularities. Previous work on learning such models has focused on methods in which the latent features are used to determine the mean and varianc...
 
Dynamical binary latent variable models for 3D human pose tracking
Found in: Computer Vision and Pattern Recognition, IEEE Computer Society Conference on
By Graham W. Taylor, Leonid Sigal, David J. Fleet, Geoffrey E. Hinton
Issue Date:June 2010
pp. 631-638
We introduce a new class of probabilistic latent variable model called the Implicit Mixture of Conditional Restricted Boltzmann Machines (imCRBM) for use in human pose tracking. Key properties of the imCRBM are as follows: (1) learning is linear in the num...
 
Distinguishing Text from Graphics in On-Line Handwritten Ink
Found in: Frontiers in Handwriting Recognition, International Workshop on
By Christopher M. Bishop, Markus Svensén, Geoffrey E. Hinton
Issue Date:October 2004
pp. 142-147
We present a system that separates text from graphics strokes in handwritten digital ink. It utilizes not just the characteristics of the strokes, but also the information provided by the gaps between the strokes, as well as the temporal characteristics of...
 
Learning Distributed Representations of Concepts Using Linear Relational Embedding
Found in: IEEE Transactions on Knowledge and Data Engineering
By Alberto Paccanaro, Geoffrey E. Hinton
Issue Date:March 2001
pp. 232-244
<p><b>Abstract</b>—In this paper, we introduce Linear Relational Embedding as a means of learning a distributed representation of concepts from data consisting of binary relations between these concepts. The key idea is to represent conce...
 
Extracting Distributed Representations of Concepts and Relations from Positive and Negative Propositions
Found in: Neural Networks, IEEE - INNS - ENNS International Joint Conference on
By Alberto Paccanaro, Geoffrey E. Hinton
Issue Date:July 2000
pp. 2259
Linear Relational Embedding (LRE) was introduced (Paccanaro and Hinton, 1999) as a means of extracting a distributed representation of concepts from relational data. The original formulation cannot use negative information and cannot properly handle data i...
 
A better way to learn features: technical perspective
Found in: Communications of the ACM
By Geoffrey E. Hinton
Issue Date:October 2011
pp. 94-94
We present a system that can reconstruct 3D geometry from large, unorganized collections of photographs such as those found by searching for a given city (e.g., Rome) on Internet photo-sharing sites. Our system is built on a set of new, distributed compute...
     
Factored conditional restricted Boltzmann Machines for modeling motion style
Found in: Proceedings of the 26th Annual International Conference on Machine Learning (ICML '09)
By Geoffrey E. Hinton, Graham W. Taylor
Issue Date:June 2009
pp. 1-8
The Conditional Restricted Boltzmann Machine (CRBM) is a recently proposed model for time series that has a rich, distributed hidden state and permits simple, exact inference. We present a new model, based on the CRBM that preserves its most important comp...
     
Keeping the neural networks simple by minimizing the description length of the weights
Found in: Proceedings of the sixth annual conference on Computational learning theory (COLT '93)
By Drew van Camp, Geoffrey E. Hinton
Issue Date:July 1993
pp. 5-13
The algorithm for pac learning k-DNF or k-CNF in the presence of malicious attribute noise in polynomial time claimed by Sloan [Slo88] does not work. It is currently open whether such an algorithm exists.
     
 1