Research map / Time series / Causal and generative time series
CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputation
YUSUKE TASHIRO, Jiaming Song, Yang Song, Stefano Ermon
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use conditional diffusion instead of unconditional diffusion | CSDI · PhysioNet 2012, 10% missing | CRPS: 0.360(0.007) -> 0.238(0.001) | Improved |
| Use conditional diffusion instead of unconditional diffusion | CSDI · PhysioNet 2012, 50% missing | CRPS: 0.458(0.008) -> 0.330(0.002) | Improved |
| Use conditional diffusion instead of unconditional diffusion | CSDI · PhysioNet 2012, 90% missing | CRPS: 0.671(0.007) -> 0.522(0.002) | Improved |
| Use conditional diffusion instead of unconditional diffusion | CSDI · Air quality | CRPS: 0.135(0.001) -> 0.108(0.001) | Improved |
| Use conditional diffusion instead of unconditional diffusion | CSDI · PhysioNet 2012, 10% missing | MAE: 0.326(0.008) -> 0.217(0.001) | Improved |
| Use conditional diffusion instead of unconditional diffusion | CSDI · PhysioNet 2012, 50% missing | MAE: 0.417(0.010) -> 0.301(0.002) | Improved |
| Use conditional diffusion instead of unconditional diffusion | CSDI · PhysioNet 2012, 90% missing | MAE: 0.625(0.010) -> 0.481(0.003) | Improved |
| Use conditional diffusion instead of unconditional diffusion | CSDI · Air quality | MAE: 12.13(0.07) -> 9.60(0.04) | Improved |
| Use CSDI rather than TimeGrad | Probabilistic forecasting · Electricity | CRPS-sum: 0.021(0.001) -> 0.017(0.000) | Improved |
| Use CSDI rather than TimeGrad | Probabilistic forecasting · Traffic | CRPS-sum: 0.044(0.006) -> 0.020(0.001) | Improved |
| Use CSDI rather than TimeGrad | Probabilistic forecasting · Taxi | CRPS-sum: 0.114(0.020) -> 0.123(0.003) | Hurt |
| Use CSDI rather than TimeGrad | Probabilistic forecasting · Wiki | CRPS-sum: 0.049(0.002) -> 0.047(0.003) | Improved |
Later papers that compared with this method
22 later accepted papers report CSDI as a baseline in their results tables; 18 of them report a better number on most of the results they share with it.
- HyFAD: Hybrid Time-Frequency Diffusion with Frequency-Aware Embedding for Time Series Imputation (NeurIPS 2026 · better on 8 of 8 shared results)
- Discretizing Continuous Time Series for Imputation with Masked Diffusion Training (NeurIPS 2026 · better on 7 of 7 shared results)
- PINFDiT: Energy-Based Physics-Informed Diffusion Transformers for General-purpose Time Series Tasks (ICLR 2026 · better on 5 of 6 shared results)
- EcoDiffusion: Uncertainty-Aware Emulation of Ecosystem Processes with Conditional Diffusion for Long Sequences with Single-Step Initialization (AAAI 2026 · better on 6 of 6 shared results)
- MMPD: Diverse Time Series Forecasting via Multi-Mode Patch Diffusion Loss (ICLR 2026 · better on 4 of 4 shared results)
- CaT-Diff: Cascaded Text-enhanced Diffusion Model for Time-Series Imputation (AAAI 2026 · better on 4 of 4 shared results)
- Time-Gated Multi-Scale Flow Matching for Time-Series Imputation (ICLR 2026 · better on 3 of 3 shared results)
- T1: One-to-One Channel-Head Binding for Multivariate Time-Series Imputation (ICLR 2026 · better on 2 of 2 shared results)
14 more not listed.
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BRITS — BRITS: Bidirectional Recurrent Imputation for Time Series (NeurIPS 2018) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Diffusion Transformers for Imputation: Statistical Efficiency and Uncertainty Quantification (NeurIPS 2025 · similarity 0.93)
- Discretizing Continuous Time Series for Imputation with Masked Diffusion Training (NeurIPS 2026 · similarity 0.90)
- ProCTI: Prototype-Refined Global Conditioning for Diffusion-Based Time Series Imputation (NeurIPS 2026 · similarity 0.89)
- Rethinking the Diffusion Models for Missing Data Imputation: A Gradient Flow Perspective (NeurIPS 2024 · similarity 0.88)
- BayOTIDE: Bayesian Online Multivariate Time Series Imputation with Functional Decomposition (ICML 2024 · similarity 0.87)
- Frequency-aware Generative Models for Multivariate Time Series Imputation (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
