Related Experiment Video
Updated: Jan 18, 2026
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.7K
Structure Learning of Deep Gaussian and Non-Gaussian Information Fusion Framework for Automated Predictive Data
IEEE Transactions on Cybernetics
|September 9, 2025
Summary
This study introduces automated structure learning for deep learning models using the maximal information coefficient (MIC). This approach enhances data-driven analytics and improves online prediction in dynamic industrial settings.
Area of Science:
- Artificial Intelligence
- Machine Learning
- Data Science
Background:
- Deep learning models combine Gaussian and non-Gaussian latent variables but lack automated structure learning.
- This limitation hinders automated data-driven modeling and analytics in complex environments.
Purpose of the Study:
- To develop an automated structure learning algorithm for deep latent variable models.
- To enhance data-driven modeling and analytics in time-varying industrial production.
Main Methods:
- Introduced the maximal information coefficient (MIC) to measure latent variable associations.
- Defined an evaluation index for automatic determination of hidden layers during model training.
- Assessed model structure necessity by evaluating the addition of new hidden layers.
Main Results:
- The proposed structure learning algorithm is feasible and effective.
- Automated data analytics significantly improved online prediction performance.
- Demonstrated success in two real-world industrial examples.
Conclusions:
- Automated structure learning addresses a critical gap in deep latent variable models.
- The developed scheme enhances efficiency and prediction accuracy in dynamic industrial processes.
