Bayesian Backpropagation Over I-O Functions Rather Than Weights

By David Hilton Wolpert. 1993

How researchers see this paper

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

1. J. Urgen Schmidhuber
2. A. Bronstein
3. Chunyuan Li
4. Yoshua Bengio
5. Sergey Levine
6. Sepp Hochreiter
7. Quoc V. Le
8. Yi Yang

Citations in the corpus

• Deep learning in neural networks: An overview (2014, background): “See also Amari and Murata (1993), Guyon, Vapnik, Boser, Bottou, and Solla (1992), Holden (1994), Vapnik (1992), Wang, Venkatesh, and Judd (1994) and Wolpert (1994).”
• On Learning to Think: Algorithmic Information Theory for Novel Combinations of Reinforcement Learning Controllers and Recurrent Neural World Models (2015, background): “2 M ’s Training To decrease bitsM + bitsH , we add a regularizing term to E, to punish excessive complexity [4, 5, 91, 294, 155, 135, 97, 170, 169, 104, 290, 7, 290, 87, 286, 319, 100, 102].”
• Linearly Combining Density Estimators via Stacking (1999, unknown): “See (Wolpert & Wolf, 1995) and (Wolpert, 1994).”
• Flat Minima (1997, methodology): “Unlike Wolpert's approach, ours distinguishes the probabilities of targets f = net(w) with det net 0 (w) = 0.”
• Loss aware post-training quantization (2019, background): “Flat minimum allow lower precision of parameters [12, 23] and are more robust under perturbations.”