We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Functional Analysis

Title: A Solution for the Greedy Approximation of a Step Function with a Waveform Dictionary

Abstract: In this paper we consider a step function characterized by an arbitrary sequence of real-valued scalars and approximate it with a matching pursuit algorithm by utilizing a waveform dictionary with rectangular window functions. This approximation problem is motivated by the need to transform noisy signals or data into an approachable form. We show that the waveform dictionary is not necessary when the scalars are nonnegative and the parameters of a wavelet dictionary on an integer lattice yields a closed-form solution for the optimization problem as part of the matching pursuit algorithm. Additionally, for any real-valued scalar sequence that may contain negative terms, we provide a suboptimal solution that mimics the closed-form solution of the nonnegative case at each iteration of the algorithm. This allows for practical computations of the approximating function, which we use to provide examples on univariate time series data.
Subjects: Functional Analysis (math.FA)
MSC classes: 42C40, 65T60
Cite as: arXiv:2206.11158 [math.FA]
  (or arXiv:2206.11158v1 [math.FA] for this version)

Submission history

From: Pierluigi Vellucci [view email]
[v1] Wed, 22 Jun 2022 15:06:20 GMT (564kb,D)
[v2] Thu, 8 Sep 2022 08:49:27 GMT (935kb,D)

Link back to: arXiv, form interface, contact.