Dynamic Graph Structure Estimation for Learning Multivariate Point Process using Spiking Neural Networks
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913771866292224 |
|---|---|
| author | Chakraborty, Biswadeep Kumawat, Hemant Kang, Beomseok Mukhopadhyay, Saibal |
| author_facet | Chakraborty, Biswadeep Kumawat, Hemant Kang, Beomseok Mukhopadhyay, Saibal |
| contents | Modeling and predicting temporal point processes (TPPs) is critical in domains such as neuroscience, epidemiology, finance, and social sciences. We introduce the Spiking Dynamic Graph Network (SDGN), a novel framework that leverages the temporal processing capabilities of spiking neural networks (SNNs) and spike-timing-dependent plasticity (STDP) to dynamically estimate underlying spatio-temporal functional graphs. Unlike existing methods that rely on predefined or static graph structures, SDGN adapts to any dataset by learning dynamic spatio-temporal dependencies directly from the event data, enhancing generalizability and robustness. While SDGN offers significant improvements over prior methods, we acknowledge its limitations in handling dense graphs and certain non-Gaussian dependencies, providing opportunities for future refinement. Our evaluations, conducted on both synthetic and real-world datasets including NYC Taxi, 911, Reddit, and Stack Overflow, demonstrate that SDGN achieves superior predictive accuracy while maintaining computational efficiency. Furthermore, we include ablation studies to highlight the contributions of its core components. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_01246 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Dynamic Graph Structure Estimation for Learning Multivariate Point Process using Spiking Neural Networks Chakraborty, Biswadeep Kumawat, Hemant Kang, Beomseok Mukhopadhyay, Saibal Machine Learning Artificial Intelligence Modeling and predicting temporal point processes (TPPs) is critical in domains such as neuroscience, epidemiology, finance, and social sciences. We introduce the Spiking Dynamic Graph Network (SDGN), a novel framework that leverages the temporal processing capabilities of spiking neural networks (SNNs) and spike-timing-dependent plasticity (STDP) to dynamically estimate underlying spatio-temporal functional graphs. Unlike existing methods that rely on predefined or static graph structures, SDGN adapts to any dataset by learning dynamic spatio-temporal dependencies directly from the event data, enhancing generalizability and robustness. While SDGN offers significant improvements over prior methods, we acknowledge its limitations in handling dense graphs and certain non-Gaussian dependencies, providing opportunities for future refinement. Our evaluations, conducted on both synthetic and real-world datasets including NYC Taxi, 911, Reddit, and Stack Overflow, demonstrate that SDGN achieves superior predictive accuracy while maintaining computational efficiency. Furthermore, we include ablation studies to highlight the contributions of its core components. |
| title | Dynamic Graph Structure Estimation for Learning Multivariate Point Process using Spiking Neural Networks |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2504.01246 |