J. O'Brien (1998). An Algorithm for the Fusion of Correlated Probabilities. Procs. of FUSION'98, Las Vegas, July 1998. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
Learning Fixed-dimension Linear Thresholds From Fragmented Data - Goldberg (1999) (Correct)
....distributions) is a distinctive novel feature of this computational learning problem, with a lot of theoretical interest. By sample complexity we mean the number of examples needed for PAC learning by a computationally unbounded learner. Experimental work in the data fusion literature such as [12, 18] has shown the strong e ect that varying assumptions about the input distribution may have on predictive performance. We aim to provide some theoretical explanation by identifying features of an input distribution that make it helpful and give associated sample size bounds. We mention ....
....to learn in the sense of [10] for learning in situations where points near the boundary may be mislabeled. For practical purposes we would like to extend these results to deal with the presence of other models of class overlap besides just uniform misclassi cation noise. The experimental work of [12, 18] assumes members of di erent classes are generated by separate Gaussian sources, and seeks the best linear threshold (minimum misclassi cation rate) There are also many possible extensions to other stochastic missing data mechanisms, which may be of practical importance while invalidating the ....
J. O'Brien (1998). An Algorithm for the Fusion of Correlated Probabilities. Procs. of FUSION'98, Las Vegas, July 1998.
Learning Fixed-dimension Linear Thresholds From Fragmented Data - Goldberg (1999) (Correct)
....It is known from this work that it is necessary to already have a lot of knowledge of the input distribution, in order to learn the function. We focus on the question of which distributions are helpful or unhelpful for 1 RFA learning. Experimental work in the data fusion literature such as [10, 13] has shown the strong effect that varying assumptions about the input distribution may have on predictive performance. We aim to provide some theoretical explanation by identifying features of an input distribution that make it helpful and give associated sample size bounds. Note that data ....
....We may ask under what circumstances it may be possible to use the algorithm of [8] for learning in situations where points near the boundary may be mislabelled. For practical purposes we would like to extend these results to deal with the presence of class overlap. The experimental work of [10, 13] assumes members of different classes are generated by separate Gaussian sources, and seeks the best linear threshold (minimum misclassification rate) 6 ACKNOWLEDGEMENTS I would like to thank Mike Paterson for reading an earlier version of this paper and making helpful comments, John ....
J. O'Brien (1998). An Algorithm for the Fusion of Correlated Probabilities. Procs. of FUSION'98, Las Vegas, July 1998.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC