Random matrices and erasure robust frames (Q1704861)

This paper addresses the issue of robustness against erasures of finite frames. The frames considered are random frames, namely \(F=\frac{1}{\sqrt n}A\) where \(A\) is \(n\times N\) whose entries are independent Gaussian random variables of \({\mathcal N}(0,1)\) distribution. They show that such frames are robust against erasures as long as the number of the remaining vectors (vectors left after removing few columns of \(F\) ) is at least \((1+\delta_0)n\) for some \(\delta_0.\) They show that robustness with more than 50\% erasures can be achieved, in fact it can be as close as 100 \%. Since the frame consists of random numbers, the robustness against erasures is probability based. They are of very high probability. Examples are provided.