A three-domain fuzzy support vector regression for image denoising and experimental studies

Area of Science:

• Machine Learning
• Artificial Intelligence
• Signal Processing

Background:

• Traditional support vector regression (SVR) methods struggle with processing uncertain data and signals.
• Existing approaches often fail to simultaneously leverage prior knowledge and input-output data characteristics.
• Enhancing the capabilities of two-domain SVR (2DSVR) for complex data analysis remains a challenge.

Purpose of the Study:

• To introduce a novel three-domain fuzzy support vector regression (3DFSVR) model.
• To develop a three-domain fuzzy kernel function (3DFKF) capable of processing uncertainties and input-output data simultaneously.
• To enhance the potential of 2DSVR by incorporating prior knowledge through a fuzzy domain for uncertain data analysis.

Main Methods:

• The proposed 3DFSVR integrates a novel three-domain fuzzy kernel function (3DFKF).
• The 3DFKF combines kernel and fuzzy membership functions into a unified three-domain function.
• A theoretical framework is constructed through the definition and solution of a fuzzy convex optimization problem.

Main Results:

• Experimental and simulation results demonstrate the effectiveness of the 3DFSVR model.
• The 3DFSVR approach shows significant improvements in uncertain image denoising tasks.
• The integration of fuzzy domains allows for superior analysis of uncertain data and signals compared to traditional methods.

Conclusions:

• The novel 3DFSVR model offers a robust solution for processing uncertain data.
• The 3DFKF effectively handles uncertainties and input-output data information simultaneously.
• 3DFSVR significantly enhances the performance of image denoising by leveraging prior knowledge and fuzzy analysis.