Exponential Bounds of Mean Error for the Kernal Estimates of Regression Functions.
Abstract
Let (X,Y), (X sub 1, Y sub 1),...,(X sub n, Y sub n) be i.i.d. R sub L x R-valued random vectors with E/Y/e infinity, and let Q sub n be a kernel estimate of the regression function Q(x) = E(Y/X = x). This paper establishes an exponential bound of the mean deviation between Q sub n and Q(x) given the training sample Z sub n = (X sub 1, Y sub 1), ...,X sub n, Y sub n), under the conditions as weak as possible. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1985
- Accession Number
- ADA167345
Entities
People
- L. C. Zhao
Organizations
- University of Pittsburgh