Network data, which captures relationships and interactions among entities, is central to many modern AI and machine learning applications in areas such as neuroscience, social science, economics, and biomedicine. Examples include brain connectivity networks, social interaction graphs, and recommendation systems. This project develops new machine learning and statistical methods for analyzing complex network data, with a focus on prediction, representation learning, and comparing populations of networks. The project emphasizes interpretable and reliable AI methods that quantify uncertainty, provide theoretical performance guarantees, and adapt to heterogeneity across nodes, individuals, and networks. The project will also contribute to training graduate students in AI, machine learning, and network data science. Technically, the project focuses on developing AI-driven and statistically rigorous methods for heterogeneous network analysis. For single-network settings, it will develop tools to quantify the predictive contributions of both node covariates and network structure within flexible black-box machine learning models, including modern AI approaches. The project will also design methods to detect and test for latent subgroups in network-linked data that differ from the broader population. For multiple-network analysis, the project will develop methodologies to test differences between collections of networks and to estimate shared, group-specific, and individual-level network structure across populations. Applications include comparing brain connectivity networks between patient and control groups, both globally and within localized subnetworks such as specific brain regions. Across these problems, the project combines network machine learning, uncertainty quantification, and scalable algorithm design with rigorous theoretical guarantees. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundati