Główne pojęcia
C2TSDは、時間的な分離と対照学習を組み合わせた条件付き拡散フレームワークであり、優れた性能を示します。
Statystyki
"Diffusion models offer a solution to the error accumulation problem—they gradually add random noise to data and learn to remove the noise in a reverse process while estimating missing values."
"Our model has demonstrated superior performance to a number of state-of-the-art baselines on various datasets under different missing patterns of spatiotemporal data."
Cytaty
"Generative models, especially diffusion models, can potentially circumvent the reliance on inaccurate, previously imputed values for future predictions."
"We propose C2TSD, a contrastive diffusion framework for spatiotemporal imputation, which constructs and uses conditional information with disentangled temporal representations and spatial relationships."