Continual learning classification method with human-in-the-loop.

Area of Science:

• Machine Learning
• Artificial Intelligence
• Computational Intelligence

Background:

• Classification is crucial for machine learning tasks like fault detection and behavior recognition.
• Continual learning algorithms address challenges with new data but often lack feedback, leading to slow convergence and potential errors.
• Existing methods struggle with efficient adaptation to evolving datasets without complete retraining.

Purpose of the Study:

• To propose a novel continual learning classification method incorporating human-in-the-loop (H-CLCM) for improved accuracy and efficiency.
• To leverage artificial immune system principles to enhance the model's adaptive learning capabilities.
• To enable the classification model to learn new data categories without requiring full retraining.

Main Methods:

• Developed a continual learning classification method (H-CLCM) inspired by artificial immune systems.
• Integrated human intervention into the testing phase for supervisory guidance.
• Implemented online parameter adjustment based on identified errors and human feedback.

Main Results:

• H-CLCM converges to accurate prediction models with reduced computational cost.
• The method effectively integrates human experience to enhance classification performance.
• The model demonstrates the ability to recognize and learn new data categories.

Conclusions:

• H-CLCM offers an effective approach to continual learning classification by incorporating human expertise.
• The integration of human-in-the-loop significantly improves model adaptability and accuracy.
• This method provides a cost-effective solution for dynamic classification tasks requiring adaptation to new data.