Using a dynamic arithmetic optimization approach to improve ridgelet neural network performance in remote sensing...

Area of Science:

• Computer Vision
• Remote Sensing
• Machine Learning

Background:

• Remote sensing scene classification is crucial for land use analysis.
• Ridgelet Neural Networks (RNNs) are effective but sensitive to hyperparameter selection.
• Optimizing hyperparameters is key to enhancing RNN performance in image processing.

Purpose of the Study:

• To propose an innovative methodology for accurate remote sensing scene classification.
• To introduce a Dynamic Arithmetic Optimization Algorithm (DAOA) for RNN hyperparameter optimization.
• To enhance the efficiency and precision of RNN models in remote sensing applications.

Main Methods:

• Developed a novel Dynamic Arithmetic Optimization Algorithm (DAOA).
• Utilized DAOA to automatically search for optimal hyperparameters for the Ridgelet Neural Network (RNN).
• Evaluated the proposed method on the UC Merced Land Use dataset.

Main Results:

• The proposed DAOA significantly enhanced the performance of the RNN model.
• Achieved superior efficiency and accuracy compared to existing state-of-the-art methods.
• Demonstrated the effectiveness of automated hyperparameter optimization for remote sensing tasks.

Conclusions:

• The combination of DAOA and RNNs offers a powerful approach for remote sensing scene classification.
• Automated hyperparameter optimization is vital for maximizing the potential of deep learning models.
• This methodology provides a more precise and efficient solution for analyzing remote sensing imagery.