Primula-3for Probabilistic Modeling and Reasoning on Graph Data

Primula-3 is a powerful software for probabilistic modeling and reasoning on graph data, combining relational Bayesian networks (RBNs) with graph neural networks (GNNs) implemented in PyTorch. This integration enables neuro-symbolic reasoning on graph-structured data, making it a versatile tool for various graph analysis tasks. The software supports a wide range of functionalities, including predictive inference, general probabilistic reasoning, and unsupervised learning. Primula-3 is designed for research and educational purposes, providing a uniform framework for modeling and learning with graph data. The text provides a detailed overview of the software's capabilities, including its support for categorical variables and its integration with PyTorch. It also offers practical examples and tutorials, demonstrating the software's application in areas such as environmental modeling, information diffusion, and community detection. The text concludes with a discussion of related tools and the unique features of Primula-3, making it a valuable resource for professionals in the field of artificial intelligence and data analysis.