Fast Algorithms for Delta-Separated Sparsity Projection

We describe a fast approximation algorithm for the ∆-separated sparsity projection problem. The ∆-separated sparsity model was introduced by Hegde, Duarte and Cevher (2009) to capture the firing process of a single Poisson neuron with absolute refractoriness. The running time of our projection algorithm is linear for an arbitrary (but fixed) precision and it is both a head and a tail approximation. This solves a problem of Hegde, Indyk and Schmidt (2015). We also describe how our algorithm fits into the approximate model iterative hard tresholding framework of Hegde, Indyk and Schmidt (2014) that allows to recover ∆-separated sparse signals from noisy random linear measurements. The resulting recovery algorithm is substantially faster than the existing one, at least for large data sets.