J. Platero Puig, J. Mateu Mahiques
Point process models are fundamental tools for analyzing spatial and temporal data, with widespread applications in many fields. Estimating likelihood functions and parameters in these processes is a critical but challenging task, especially when dealing with complex or high-dimensional data. This study introduces a framework that leverages neural network architectures to address these challenges. Specifically, we examine the potential of Convolutional Neural Networks (CNNs) and Graph Neural Networks (GNNs) for effectively approximating likelihood functions and estimating parameters in point process models. Our approach takes advantage of the expressive capabilities of these networks to model the intricate dependencies and spatial-temporal patterns characteristic of point processes. Initial findings indicate that neural network-based methods offer competitive advantages over traditional statistical estimation techniques, particularly in terms of scalability and adaptability.
Keywords: Point processes, Neural Networks, likelihood free inference
Scheduled
Spatio-Temporal Statistics I
June 11, 2025 3:30 PM
MR 3