Unsupervised Learning Procedures for Neural Networks

By Suzanna Becker. 1991

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Deep learning in neural networks: An overview (2014, background): “Such codes may be sparse and can be advantageous for (1) data compression, (2) speeding up subsequent BP (Becker, 1991), (3) trivializing the task of subsequent naive yet optimal Bayes classifiers (Schmidhuber et al., 1996).”
Learning Factorial Codes by Predictability Minimization (1992, background): “As Becker (1991) observes, if a representation with uncorrelated components is used as the input to a higher level linear supervised learning network, then the Hessian of the supervised network's error function is diagonal, thus allowing efficient methods for speeding up learning (note that statisti”
A Survey of Inductive Biases for Factorial Representation-Learning (2016, methodology): “More generally, they are useful as features used as input to supervised learning methods, as observed by [8] for linear supervised tasks.”
Feature Extraction Through LOCOCODE (1999, background): “Such codes can be advantageous for data compression, speeding up subsequent gradient-descent learning (e.g., Becker, 1991), and simplifying subsequent Bayes classifiers (e.g., Schmidhuber, Eldracher, & Foltin, 1996).”
A Transdisciplinary Review of Deep Learning Research and Its Relevance for Water Resources Scientists (2017, background): “, the Kullback-Leibler divergence (a measure of the difference between statistical distributions) [Kullback and Leibler, 1951] or information entropy [Shannon, 1948], and may extract invariant features from diverse data points [Becker, 1991].”