Weiguo Lu

Postdoc researcher

Email：luweiguo@@gdiist.cn

Personal Biography:

Doctor of Mathematics from the University of Macau, with several years of experience working in the finance and venture capital industries. The focus of the doctoral research was on Gaussian Mixture Models and their related exploration in neural networks. Concepts such as Gaussian Mixture Expansion were proposed, and based on this, efficient density function approximation algorithms were developed. These algorithms significantly outperform the current mainstream Variational Bayesian methods and Expectation-Maximization algorithms. In generative models, methods such as the Gaussian Mixture Conditional Diffusion Model were proposed. A training method using the negative Gaussian mixture gradient was also introduced

Research Focus:

Currently, the main research direction continues from past studies, utilizing Gaussian Mixture Models to construct the underlying computational units of networks and develop new network modules/architectures. A preliminary design of Gaussian-like nonlinear computational units has already been developed, which has shown excellent results in initial experiments, leading to significant performance improvements and effective reduction in overfitting. On the other hand, there is an active exploration of brain-like and brain-inspired computational modules and network architectures.

Representative Publications:
[1]Lu, Weiguo, et al. "An efficient Gaussian mixture model and its application to neural networks." Knowledge-Based Systems 310 (2025): 112942.

[2]Lu, Weiguo, et al. "An effective one-iteration learning algorithm based on Gaussian mixture expansion for densities." Communications in Nonlinear Science and Numerical Simulation 142 (2025): 108494.

[3]Lu, Weiguo, et al. "Diffusion model conditioning on Gaussian mixture model and negative Gaussian mixture gradient." Neurocomputing 614 (2025): 128764.