Learning sets of filters using back-propagation

By David C. Plaut, Geoffrey E. Hinton. 1987

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

Recurrent Convolutional Neural Networks for Text Classification (2015, background): “The magnitude of the maximum or minimum equals the square root of the “fan-in”(Plaut and Hinton 1987).”
Understanding normal and impaired word reading: computational principles in quasi-regular domains. (1996, background): “…is introduced only after the first few initial epochs to avoid magnifying the effects of the initial weight gradients, which are very large because, for each word, any activity of all but a few phoneme units—those that should be active—produces a large amount of error (Plaut & Hinton, 1987).”
Neural networks for the prediction and forecasting of water resource variables: a review of modelling issues and applications (2000, background): “The advantages of larger networks are that they tend to learn quickly (in terms of number of training cycles) (Plaut and Hinton, 1987), can form adequately complex decision regions (Lippmann, 1987) and have an increased ability to avoid local minima in the error surface (Rumelhart et al., 1986b).”
Connectionist Learning Procedures (1989, methodology): “Various other methods have also been suggested (Amari, 1967: Parker, 1987; Plaut and Hinton, 1987).”
Deep Learning for Efficient Discriminative Parsing (2011, methodology): “Taskar et al. (2004) proposed a comparison setup for discriminative parsers limited to Penn Treebank sentences with ≤ 15 words.”
Discovering discrete subword units with binarized autoencoders and hidden-Markov-model encoders (2015, background): “An AE consists of an encoding and a decoding part [14].”