Netzwerkarchitekturen, Zielfunktionen und Kettenregel (Network architectures, objective functions, and chain rule)

By Jürgen Schmidhuber. 1993

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, methodology): “PM-based UL was applied not only to FNNs but also to RNNs (e.g., Lindstädt, 1993; Schmidhuber, 1993b).”
On Learning to Think: Algorithmic Information Theory for Novel Combinations of Reinforcement Learning Controllers and Recurrent Neural World Models (2015, background): “For example, deep stacks of unsupervised RNNs [228] or FNNs [13, 96, 139] help to accelerate subsequent supervised learning through BPTT [228, 230] or BP [96].”
One Big Net For Everything (2018, methodology): “It was shown that this type of unsupervised pre-training for deep learning networks can greatly facilitate the learning of additional user-deﬁned tasks [61, 64].”
Semilinear Predictability Minimization Produces Well-Known Feature Detectors (1996, background): “Theequivalence of (2) and (3) was observed by Peter Dayan, Richard Zemel and Alex Pouget (personalcommunication, SALK Institute, 1992 | see (Schmidhuber, 1993) for details).”
My First Deep Learning System of 1991 + Deep Learning Timeline 1962-2013 (2013, unknown): “Figure 1: My first Deep Learning system of 1991 used a deep stack of recurrent neural networks (a Neural Hierarchical Temporal Memory) pre-trained in unsupervised fashion to accelerate subsequent supervised learning [79, 81, 82].”
Unsupervised Pre-training of a Deep LSTM-based Stacked Autoencoder for Multivariate Time Series Forecasting Problems (2019, methodology): “The first difference concerns the objective of the prediction task of Schmidhuber’s model, which is predicting the next input from the previous inputs.”