Rethinking Time-Series Imputation as Conditional Inference along Temporal Evolution

Yu Fan, Yang Yang, Yufan Guo, Huazhong Yang, Pengjun Wang
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:28904-28931, 2026.

Abstract

Real-world time-series data often suffer from missing observations, hindering long-range temporal modeling. However, most existing imputation methods formulate imputation as conditional reconstruction over limited context, which restricts temporal information propagation and fails to explicitly model temporal evolution. To overcome this limitation, we propose the Conditional Temporal Inference Paradigm (CTIP), which formulates time-series imputation as conditional inference along temporal evolution. Under this paradigm, we introduce CBiT, which leverages a history compression mechanism to encode long-range history into a compact latent space for history-conditioned temporal imputation. In addition, we adopt a partitioned modeling strategy that distinguishes historical context and temporal imputation targets with only linear-time complexity. Extensive experiments on multiple public benchmarks show that CBiT improves imputation accuracy by reducing Masked MAE and Masked RMSE by 27.3% and 18.6%, respectively, across different missing rates.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-fan26h, title = {Rethinking Time-Series Imputation as Conditional Inference along Temporal Evolution}, author = {Fan, Yu and Yang, Yang and Guo, Yufan and Yang, Huazhong and Wang, Pengjun}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {28904--28931}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/fan26h/fan26h.pdf}, url = {https://proceedings.mlr.press/v306/fan26h.html}, abstract = {Real-world time-series data often suffer from missing observations, hindering long-range temporal modeling. However, most existing imputation methods formulate imputation as conditional reconstruction over limited context, which restricts temporal information propagation and fails to explicitly model temporal evolution. To overcome this limitation, we propose the Conditional Temporal Inference Paradigm (CTIP), which formulates time-series imputation as conditional inference along temporal evolution. Under this paradigm, we introduce CBiT, which leverages a history compression mechanism to encode long-range history into a compact latent space for history-conditioned temporal imputation. In addition, we adopt a partitioned modeling strategy that distinguishes historical context and temporal imputation targets with only linear-time complexity. Extensive experiments on multiple public benchmarks show that CBiT improves imputation accuracy by reducing Masked MAE and Masked RMSE by 27.3% and 18.6%, respectively, across different missing rates.} }
Endnote
%0 Conference Paper %T Rethinking Time-Series Imputation as Conditional Inference along Temporal Evolution %A Yu Fan %A Yang Yang %A Yufan Guo %A Huazhong Yang %A Pengjun Wang %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-fan26h %I PMLR %P 28904--28931 %U https://proceedings.mlr.press/v306/fan26h.html %V 306 %X Real-world time-series data often suffer from missing observations, hindering long-range temporal modeling. However, most existing imputation methods formulate imputation as conditional reconstruction over limited context, which restricts temporal information propagation and fails to explicitly model temporal evolution. To overcome this limitation, we propose the Conditional Temporal Inference Paradigm (CTIP), which formulates time-series imputation as conditional inference along temporal evolution. Under this paradigm, we introduce CBiT, which leverages a history compression mechanism to encode long-range history into a compact latent space for history-conditioned temporal imputation. In addition, we adopt a partitioned modeling strategy that distinguishes historical context and temporal imputation targets with only linear-time complexity. Extensive experiments on multiple public benchmarks show that CBiT improves imputation accuracy by reducing Masked MAE and Masked RMSE by 27.3% and 18.6%, respectively, across different missing rates.
APA
Fan, Y., Yang, Y., Guo, Y., Yang, H. & Wang, P.. (2026). Rethinking Time-Series Imputation as Conditional Inference along Temporal Evolution. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:28904-28931 Available from https://proceedings.mlr.press/v306/fan26h.html.

Related Material