Detail publikace

Piecewise-polynomial signal segmentation using proximal splitting convex optimization methods

Originální název

Piecewise-polynomial signal segmentation using proximal splitting convex optimization methods

Anglický název

Piecewise-polynomial signal segmentation using proximal splitting convex optimization methods

Jazyk

en

Originální abstrakt

We show how the problem of segmenting noisy piecewise polynomial signal can be formulated as a convex optimization task. Because the number of model changes in signal is considered low in comparison to the overall number of data points, we rely on the concept of sparsity and its convex-relaxed counterpart, the l1-norm. We present an unconstrained, overparametrized optimization formulation whose solution can be used for detecting the breakpoints, and for robust data denoising, in consequence. The problem is solved numerically by iterative proximal splitting methods.

Anglický abstrakt

We show how the problem of segmenting noisy piecewise polynomial signal can be formulated as a convex optimization task. Because the number of model changes in signal is considered low in comparison to the overall number of data points, we rely on the concept of sparsity and its convex-relaxed counterpart, the l1-norm. We present an unconstrained, overparametrized optimization formulation whose solution can be used for detecting the breakpoints, and for robust data denoising, in consequence. The problem is solved numerically by iterative proximal splitting methods.

Dokumenty

BibTex


@misc{BUT127474,
  author="Michaela {Novosadová} and Pavel {Rajmic}",
  title="Piecewise-polynomial signal segmentation using proximal splitting convex optimization methods",
  annote="We show how the problem of segmenting noisy piecewise polynomial signal can be formulated as a convex optimization task. Because the number of model changes in signal is considered low in comparison to the overall number of data points, we rely on the concept of sparsity and its convex-relaxed counterpart, the l1-norm. We present an unconstrained, overparametrized optimization formulation whose solution can be used for detecting the breakpoints, and for robust data denoising, in consequence. The problem is solved numerically by iterative proximal splitting methods.",
  chapter="127474",
  year="2016",
  month="june",
  pages="1--14"
}