Generalized hidden-mapping ridge regression, knowledge-leveraged inductive transfer learning for neural networks,...

Area of Science:

• Machine Learning
• Artificial Intelligence
• Data Science

Background:

• Inductive transfer learning leverages source domain data to improve target domain model training.
• Existing transfer learning methods are often model-specific, limiting their applicability.
• A need exists for flexible transfer learning approaches applicable to various classical intelligence models.

Purpose of the Study:

• To introduce a generalized transfer learning method applicable to diverse models.
• To develop an inductive transfer learning algorithm integrating knowledge leverage with GHRR.
• To evaluate the performance of the proposed TGHRR method.

Main Methods:

• Generalized Hidden-Mapping Ridge Regression (GHRR) was developed for training various models.
• A knowledge-leverage based transfer learning mechanism was integrated with GHRR, creating TGHRR.
• TGHRR utilizes induced knowledge for clearer and more concise data distribution control.

Main Results:

• Experimental evaluation included regression and classification on synthetic and real-world datasets.
• The proposed GHRR and TGHRR algorithms demonstrated effectiveness.
• TGHRR performance was competitive with or superior to state-of-the-art inductive transfer learning algorithms.

Conclusions:

• TGHRR offers a flexible and effective inductive transfer learning solution.
• The method successfully leverages source domain knowledge for improved target domain model performance.
• TGHRR provides a viable alternative to model-specific transfer learning techniques.