[Home]
[
News]
[ Events]
[ People]
[ Research]
[ Education]
[Visitor Info]
[UCSD Only]
[Admin]
Privacy-Preserving Linear and Nonlinear Approximation
Olvi Mangasarian
UCSD
Abstract:
We propose a novel privacy-preserving random kernel approximation based on an m-by-n data matrix A whose rows are divided into privately held blocks. Each block of rows belongs to a different entity that is unwilling to share its rows or make them public. We wish to obtain an accurate function approximation for a given m-dimensional vector y each component of which corresponds to the m the rows of A. Our approximation to y is a real function on an n-dimensional real space and is based on the concept of a reduced kernel K(A,B′) where B′ is the transpose of a completely random matrix B. The proposed approximation, which is public but does not reveal the privately-held data matrix A, has accuracy comparable to that of an ordinary kernel approximation based on a publicly disclosed data matrix A.
Tuesday, January 24, 2012
11:00AM AP&M 2402
News]
[ Events]
[ People]
[ Research]
[ Education]
[Visitor Info]
[UCSD Only]
[Admin]
Directors:
Randolph E. Bank
Philip E. Gill
Michael Holst
Administrative Contact:
[ Click to Send Email ]
Randolph E. Bank
Philip E. Gill
Michael Holst
Administrative Contact:
[ Click to Send Email ]
Olvi Mangasarian
UCSD
Abstract:
We propose a novel privacy-preserving random kernel approximation based on an m-by-n data matrix A whose rows are divided into privately held blocks. Each block of rows belongs to a different entity that is unwilling to share its rows or make them public. We wish to obtain an accurate function approximation for a given m-dimensional vector y each component of which corresponds to the m the rows of A. Our approximation to y is a real function on an n-dimensional real space and is based on the concept of a reduced kernel K(A,B′) where B′ is the transpose of a completely random matrix B. The proposed approximation, which is public but does not reveal the privately-held data matrix A, has accuracy comparable to that of an ordinary kernel approximation based on a publicly disclosed data matrix A.
Tuesday, January 24, 2012
11:00AM AP&M 2402