This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Week 7 Lecture 12 Fourier estimation and Linear Minimaxity An orthonormal basis for L 2 ([0 ; 1]) is & 1 ( x ) = 1 & 2 k ( x ) = p 2 cos (2 kt ) ;& 2 k ( x ) = p 2 sin (2 kx ) ; k & 1 . The periodic Sobolev class f W & 2 ( M ) is de&ned as F = & f : Z 1 f ( m ) 2 M; f ( j ) (0) = f ( j ) (1) , j = 0 ;:::;m 1 Let j = f;& j . It is known that F = ( f : 1 X k =1 (2 k ) 2 m 2 2 k + 2 2 k +1 M ) . Why? For any f 2 F , we have f ( m ) ( x ) = (2 k ) m 1 X k =1 h 2 k p 2 cos (2 kt ) + 2 k +1 p 2 sin (2 kt ) i , m & 1 integers for m = 0 ; 1 ;:::; 1 . Apparently f ( m ) (0) = f ( m ) (1) , m = 0 , 1 , ::: , 1 , and R f ( & ) 2 M . Remark : The Sobolev class W m 2 ( M ) is de&ned as F = & f : Z 1 f ( m ) 2 M (without periodic boundary conditions). The representation of the model in the ellipsoid form can be achieved using the DemmlerReinsch spline. For the model dy ( t ) = f ( t ) dt + dW ( t ) , = n & 1 = 2 or an orthogonal transformation of the white noise model, Z & j dy ( t ) = Z & j f ( t ) dt + Z & j dW ( t ) i.e., y i...
View
Full
Document
 '09
 HarrisonH.Zhou

Click to edit the document details