Competitive On‐line Statistics

By Vladimir G Vovk. 2001

311 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Management Science and Operations Research

View Paper: https://www.semanticscholar.org/paper/11335fd1a9e28c00b2e92c1be97b0a81b62e4ffa

How researchers see this paper

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

In Computer Networks and Communications

In Management Science and Operations Research

Citations in the corpus

Online Bounds for Bayesian Algorithms (2004, methodology): “Vovk (2001) provides such an algorithm, though this algorithm differs from MAP in that its predictions at timet are a nonlinear function ofx(t) (it usesAt instead of At−1 at time t).”
Theory and applications of proper scoring rules (2014, unknown): “The aggregating algorithm of [28] may supply a path ahead.”
General Loss Bounds for Universal Sequence Prediction (2001, background): “There are good introductions and surveys of Solomono sequence prediction [LV97], inductive inference [AS83, Sol97], reasoning under uncertainty [Gr¨u98], and competitive online statistics [ Vov99 ] with interesting relations to this work.”
Using Stacking to Average Bayesian Predictive Distributions (with Discussion) (2017, background): “…notation transposes the arguments of d compared with that of the paper. statisticians), developed in the discipline of “prediction with expert advice”(Vovk, 2001), which shows that, under suitable conditions, we can construct a prequential strategy P which for any data whatsoever will perform…”
A Closer Look at Adaptive Regret (2012, background): “We discuss both and show that each of them yields Fixed Share with a particular choice of time-dependent switching rate α t .”
Aggregation by exponential weighting, sharp PAC-Bayesian bounds and sparsity (2008, background): “Note that, for bounded errors ξi , oracle inequalities of the form (16) are also established in the theory of prediction of deterministic sequences (Vovk 1990; Littlestone and Warmuth 1994; Cesa-Bianchi et al. 1997; Kivinen and Warmuth 1999; Cesa-Bianchi and Lugosi 2006).”
Exact minimax risk for linear least squares, and the lower tail of sample covariance matrices (2019, background): “Perhaps most different from the point of view adopted here is the approach from [28, 76, 6, 61, 8], whose authors consider worst-case covariates (either in the individual sequences or in the agnostic learning setting) under boundedness assumptions on both covariates and outputs, and investigate achi”
Learning by mirror averaging (2005, methodology): “Interestingly, for such problems, which are quite diﬀerent from ours, methods similar to (3.3) constitute one of the principal tools; cf. [11, 13, 23, 26, 27].”
Prediction, learning, and games (2006, methodology): “5 was introduced by Vovk [298, 300], and also studied by Haussler, Kivinen, and Warmuth [151], who characterize the loss functions for which (ln N ) regret bounds are possible.”
Kernel Ridge Regression (2013, background): “7) with the coefficient less than 1 instead of 4, even in the case where kxtk1 X for all t : see Theorem 2 in [16].”