Heteroscedastic Sequences: Beyond Gaussianity
نویسندگان
چکیده
We address the problem of sequential prediction in the heteroscedastic setting, when both the signal and its variance are assumed to depend on explanatory variables. By applying regret minimization techniques, we devise an efficient online learning algorithm for the problem, without assuming that the error terms comply with a specific distribution. We show that our algorithm can be adjusted to provide confidence bounds for its predictions, and provide an application to ARCH models. The theoretical results are corroborated by an empirical study.
منابع مشابه
ar X iv : a st ro - p h / 06 04 29 8 v 1 1 3 A pr 2 00 6 the Non - Gaussianity of Racetrack Inflation Models
In this paper, we use the result in [7] to calculate the non-Gaussianity of the racetrack models in [3, 5]. Our results show that the nonGaussianity in [3] is reasonable, while the non-Gaussianity in [5] is beyond the observational limits.
متن کاملar X iv : a st ro - p h / 06 04 29 8 v 2 1 4 A pr 2 00 6 the Non - Gaussianity of Racetrack Inflation Models
In this paper, we use the result in [7] to calculate the non-Gaussianity of the racetrack models in [3, 5]. Our results show that the nonGaussianity in [3] is reasonable, while the non-Gaussianity in [5] is beyond the observational limits. PACS: 98.80.Cq, 98.80.Jk, 04.20.Gz
متن کاملSpectral Running and Non-Gaussianity from Slow-Roll Inflation in Generalised Two–Field Models
Theories beyond the standard model such as string theory motivate low energy effective field theories with several scalar fields which are not only coupled through a potential but also through their kinetic terms. For such theories we derive the general formulae for the running of the spectral indices for the adiabatic, isocurvature and correlation spectra in the case of two field inflation. We...
متن کاملThree laws of feedback systems: entropy rate never decreases, generalized Bode integral, absolute lower bound in variance minimization, Gaussianity-whiteness measure (joint Shannon-Wiener entropy), Gaussianing-whitening control, and beyond
This paper aims at obtaining universal laws and absolute lower bounds of feedback systems using information theory. The feedback system setup is that with causal plants and causal controllers. Three laws (entropy rate never decreases, generalized Bode integral, and absolute lower bound in variance minimization) are obtained, which are in entropy domain, frequency domain, and time domain, respec...
متن کاملEmpirical estimates for various correlations in longitudinal-dynamic heteroscedastic hierarchical normal models
In this paper, we first define longitudinal-dynamic heteroscedastic hierarchical normal models. These models can be used to fit longitudinal data in which the dependency structure is constructed through a dynamic model rather than observations. We discuss different methods for estimating the hyper-parameters. Then the corresponding estimates for the hyper-parameter that causes the association...
متن کامل