| Bartlett, P. L., Long, P. M., and Williamson, R. C. (1994). |
....the random variable z, is the weighted sum of N iid gaussian variables, so it is also a gaussian, with variance oe 2 z ; the second term is just V in the noise free case. So how does this effect the VC dimension Noise in this sense does not fall into the classical VC framework (although Bartlett, Long, and Williamson (1994) extend the framework towards situations of this sort. Nevertheless, the effect is clear: there is some probability Pm of misclassifying each input. This probability depends on V and on z: it is the probability that sgn( V z) 6= sgn( V ) If both z and V are 0 mean, then this is ....
Bartlett, P. L., Long, P. M., and Williamson, R. C. (1994).
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