Sanov Property, Generalized $I$-Projection and a Conditional Limit Theorem
By Imre Csiszár. 1984
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
- Yoshua Bengio
- Percy Liang
- Tong Zhang
- Dacheng Tao
- Ruslan Salakhutdinov
- Sergey Levine
- Trevor Darrell
- Quoc V. Le
Citations in the corpus
- CPR: Classifier-Projection Regularization for Continual Learning (2020, methodology): “Furthermore, we show in the next section that the KL-term can be geometrically interpreted in terms of information projections (Csiszár, 1984; Cover & Thomas, 2012; Murphy, 2012), providing an additional argument (besides promoting wide local minima) for the benefit of using CPR in continual learnin”
- A large deviation approach to some transportation cost inequalities (2005, background): “In other words, µA is Csisz´ar’s I-projection of µ on { 2 P(X) : (A) = 1}, see [5, 6 ]. If is such that (A) = 1, one has”
- Nonequilibrium Markov Processes Conditioned on Large Deviations (2014, background): “where k is a real parameter related to the conditioning value s and W (k) is the generating function of P (x) normalizing the so-called exponentially tilted distribution Pk(x); see [17–20] for details.”
- Information and Entropy Econometrics—Editor's View (2002, background): “For a detailed discussion, see the classic text by Ellis (1985) and the early work of Csiszar (1984).”
- Some inequalities for information divergence and related measures of discrimination (2000, methodology): “For more specific applications of Pinsker’s inequality we can, for instance, point to Csiszar [ 11 ], Marton [28], [29], Ryabko and Topsoe [35], and Topsoe [36] for five different applications.”
- APPROXIMATION OF DENSITY FUNCTIONS BY SEQUENCES OF EXPONENTIAL FAMILIES (1991, background): “The conditional limit characterization of Van Campenhout and Cover (1981) and Csiszar (1984) establishes that for large n, Pn m is the asymptotic conditional probability density function for X1 given (1/n)Ei=l4k(Xi) = ak when the unconditional density is taken to be pO.”
- Information and Entropy Econometrics - A Review and Synthesis (2008, background): “set of axioms is discussed in Csiszar and Korner (1981). However, as noted by Shannon, “ .”
- Connections Between Entropic and Linear Projections in Asset Pricing Estimation (2002, unknown): “A closely related large deviations interpretation of the Gibbs canonical measure is given by Conditional limit theorems, e.g. Cover and Van Campenhout (1981) and Csiszar (1984).”
- Logistic Regression, AdaBoost and Bregman Distances (2000, background): “The result appears to be due to Csisz´ar [6, 7] and Topsoe [21].”