Modified Noise-Immune Fuzzy Neural Network for Solving the Quadratic Programming With Equality Constraint Problem.

Area of Science:

• Optimization
• Artificial Intelligence
• Control Systems

Background:

• Quadratic programming with equality constraint (QPEC) problems are widely used in various industries.
• Noise interference poses a significant challenge in solving QPEC problems in real-world applications.
• Existing recurrent neural network models often struggle with noise interference.

Purpose of the Study:

• To propose a novel Modified Noise-Immune Fuzzy Neural Network (MNIFNN) model.
• To enhance the noise tolerance and robustness of neural network-based QPEC solvers.
• To improve the adaptability of the QPEC solving model through fuzzy logic.

Main Methods:

• Development of a Modified Noise-Immune Fuzzy Neural Network (MNIFNN).
• Integration of proportional, integral, and differential elements for enhanced robustness.
• Utilization of two fuzzy logic systems (FLSs) to generate adaptive design parameters based on residual and residual integral terms.

Main Results:

• The proposed MNIFNN model exhibits inherent noise tolerance.
• The MNIFNN demonstrates superior robustness compared to Traditional Gradient Recurrent Neural Network (TGRNN) and Traditional Zeroing Recurrent Neural Network (TZRNN) models.
• Numerical simulations confirm the effectiveness of the MNIFNN in handling noise interference.

Conclusions:

• The MNIFNN is an effective and robust model for solving QPEC problems, especially in noisy environments.
• The adaptive fuzzy parameter design enhances the model's applicability.
• The MNIFNN offers a promising alternative to existing methods for noise-sensitive QPEC applications.