STKAN architecture enhances spatio-temporal forecasting with Kolmogorov-Arnold Networks

Researchers have introduced STKAN, a novel architecture for spatio-temporal forecasting that integrates Taylor-polynomial Kolmogorov-Arnold Network modules. This approach aims to improve the modeling of complex real-world data, such as traffic patterns, which exhibit both spatial correlations and nonlinear temporal dynamics. STKAN constructs spatial representations through a soft node-group assignment mechanism and then models temporal dependencies, incorporating self-attention layers for long-range interactions. Experiments on five traffic forecasting benchmarks indicate that STKAN performs competitively and outperforms an MLP-based variant. AI

IMPACT This research could lead to more accurate forecasting models for complex spatio-temporal data, impacting fields like traffic management and climate modeling.

RANK_REASON The cluster describes a new research paper introducing a novel architecture for spatio-temporal forecasting. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• IArxiv
• Kolmogorov--Arnold Networks
• MLP-based variant
• ScienceCast
• STKAN
• Taylor-polynomial Kolmogorov-Arnold Network

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →