Summary: Variants of known results New results
Risk bounds for linear regression
Jean-Yves Audibert1,2 & Olivier Catoni3
1. Imagine - Université Paris Est,
2. Willow - CNRS/ENS/INRIA
3. DMA - CNRS/ENS
May 2009
Variants of known results New results
Ordinary least squares and empirical risk minimization
Linear aggregation: F = Flin = span{1, . . . , d } and f
lin = f
Let ^f(ols)
argminfFlin
1
n
n
i=1[Yi - f(Xi )]2
.
ER(^f(ols)
) - R(f

Source: Audibert, Jean-Yves - Département d'Informatique, École Normale Supérieure

Collections: Computer Technologies and Information Sciences