Time-series estimation from randomly time-warped observations

• We consider the problem of estimating a signal given randomly time warped observations signal. Many existing methods rely on unknown warps, computationally challenging problem. propose scalable method based interpreting as samples manifold. search for ’center’ manifold, by resorting to techniques from graph/network processing. demonstrate that number required is not exceedingly high practical scenarios. its observations. Such estimation commonly performed altering through some inverse-warping, or solving demanding optimization formulation . While these may be unavoidable if are few, when large amounts available, cost running such algorithms can prohibitive. scenario where we have many observations, and simple algorithm function interest. utility streaming biomedical signals.