title: Sparse Regression Codes for Lossy Compression

abstract: We study the lossy compression performance of Sparse
Regression Codes.
These are codes where each codeword is a sparse combination of columns
of a design
matrix. For an IID Gaussian source with minimum-distance decoding and
squared-error
distortion it is shown that these codes achieve the Shannon
rate-distortion bound.
The proof is based on a refinement of the second moment method.

This is joint work with Ramji Venkataramanan.