Multivariate time-series imputation with disentangled temporal representations.

Shuai Liu, Xiucheng Li, Gao Cong, Yile Chen, Yue Jiang.

The paper “Multivariate Time-Series Imputation with Disentangled Temporal Representations” proposes TIDER, a scalable matrix-factorization based method that decomposes time series into disentangled components (trend, seasonality, bias) for more interpretable and accurate imputation.