Estimating the mean and variance of the target probability distribution
作者： D.A. Nix , A.S. Weigend
关键词: Inverse-chi-squared distribution 、 Normal-gamma distribution 、 Compound probability distribution 、 Statistics 、 Mathematics 、 Half-normal distribution 、 Applied mathematics 、 Student's t-distribution 、 Probability distribution 、 Rayleigh distribution 、 Joint probability distribution
摘要: Introduces a method that estimates the mean and variance of probability distribution target as function input, given an assumed error-distribution model. Through activation auxiliary output unit, this provides measure uncertainty usual network for each input pattern. The authors derive cost weight-update equations example Gaussian error distribution, demonstrate feasibility on synthetic problem where true input-dependent noise level is known. >
sci-hub.se 参考文章(4)
Andreas S. Weigend, Bernardo A. Huberman, David E. Rumelhart, PREDICTING THE FUTURE: A CONNECTIONIST APPROACH International Journal of Neural Systems. ,vol. 01, pp. 193- 209 ,(1990) , 10.1142/S0129065790000102
David J. C. MacKay, A practical Bayesian framework for backpropagation networks Neural Computation. ,vol. 4, pp. 448- 472 ,(1992) , 10.1162/NECO.1992.4.3.448
Martin Casdagli, Stephen Eubank, J.Doyne Farmer, John Gibson, State space reconstruction in the presence of noise Physica D: Nonlinear Phenomena. ,vol. 51, pp. 52- 98 ,(1991) , 10.1016/0167-2789(91)90222-U
Andreas S Weigend, None, Time Series Prediction : Forecasting The Future And Understanding The Past International Journal of Forecasting. ,vol. 10, pp. 161- 163 ,(1994) , 10.4324/9780429492648