An Inexact Penalty Decomposition Method for Sparse Optimization

Area of Science:

• Optimization
• Signal Processing
• Machine Learning

Background:

• The penalty decomposition method is a versatile technique for sparse optimization problems.
• Applications include compressed sensing, sparse logistic regression, and image restoration.
• Increasing penalty parameters can lead to time-consuming computations.

Purpose of the Study:

• To accelerate the penalty decomposition method for sparse optimization.
• To improve the efficiency of solving sparse optimization problems.

Main Methods:

• Proposes an accelerated penalty decomposition method.
• Involves finding inexact solutions to subproblems for each penalty parameter.

Main Results:

• Demonstrates effectiveness and efficiency through computational experiments.
• Accurately generates sparse and redundant representations of one-dimensional random signals.

Conclusions:

• The proposed accelerated method is effective and efficient for sparse optimization.
• Offers a faster alternative for generating sparse signal representations.