Research map / Time series / Long-term forecasting
Paloma: Phase-Conditioned Residual Modulation for Time Series Forecasting
Jingru Fei, Kun Yi, Wei Fan, Qi Zhang, Zhendong Niu
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes Paloma, which models phase-dependent structure left in time-series residuals after removing the periodic component, for time series forecasting.
Problem
Explicit periodicity modeling in time series forecasting treats the residual after removing the periodic component as unstructured noise, though the residuals may still carry structured phase-dependent deviations.
Approach
Shows that post-periodic residuals retain phase-dependent deviations and conditions residual modulation on the phase, rather than treating residuals as white noise.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Phase Rotation module (ablation) | Paloma · ETTm1, ETTh1, Electricity, Weather, Traffic, Solar-Energy | MSE: ETTm1 0.368 -> 0.363; Traffic 0.471 -> 0.461 | Improved |
| add Amplitude Scaling module (ablation) | Paloma · ETTm1, ETTh1, Electricity, Weather, Traffic, Solar-Energy | MSE: ETTm1 0.369 -> 0.363; Solar-Energy 0.201 -> 0.197 | Improved |
| add both PR and AS modules to the base cycle-residual backbone (ablation) | Paloma · ETTm1 | MSE: 0.376 -> 0.363 | Improved |
| add both PR and AS modules on ETTh1 (ablation) | Paloma · ETTh1 | MSE: w/o PR,AS 0.428 vs w/o PR 0.427, w/o AS 0.430; full 0.421 | Mixed |
| plug Paloma technique into DLinear | DLinear · PEMS03 (avg) | MSE: 0.265 -> 0.099 | Improved |
| plug Paloma technique into iTransformer | iTransformer · PEMS04 (avg) | MSE: 0.121 -> 0.092 | Improved |
| plug Paloma technique into PatchTST | PatchTST · Electricity (avg) | MSE: 0.191 -> 0.172 | Improved |
| plug Paloma technique into FilterNet | FilterNet · PEMS04 (avg) | MSE: 0.146 -> 0.113 | Improved |
| use channel-independent Paloma on Traffic with many channels | Paloma · Traffic | MSE: Paloma 0.461 vs iTransformer 0.428 | Hurt |
| assume a cycle length misaligned with the true period (ablation) | Paloma · sensitivity analysis | forecast error | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- DistDF — DistDF: Time-series Forecasting Needs Joint-distribution Wasserstein Alignment (ICLR 2026) · better on 1 of 1 shared result
- iTransformer — iTransformer: Inverted Transformers Are Effective for Time Series Forecasting (ICLR 2024) · better on 1 of 1 shared result
- TimeXer — TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variables (NeurIPS 2024) · 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.
Surprising findings
- Residuals after periodic component removal still exhibit structured, phase-dependent deviations rather than unstructured noise.
Closest papers by meaning
- PAMod: Modeling Cyclical Shifts via Phase-Amplitude Modulation for Non-stationary Time Series Forecasting (NeurIPS 2026 · similarity 0.88)
- Unlocking the Power of Patch: Patch-Based MLP for Long-Term Time Series Forecasting (AAAI 2025 · similarity 0.87)
- Tiny Time Mixers (TTMs): Fast Pre-trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate Time Series (NeurIPS 2024 · similarity 0.86)
- CycleNet: Enhancing Time Series Forecasting through Modeling Periodic Patterns (NeurIPS 2024 · similarity 0.86)
- xLSTM-Mixer: Multivariate Time Series Forecasting by Mixing via Scalar Memories (NeurIPS 2025 · similarity 0.86)
- TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting (NeurIPS 2025 · similarity 0.86)
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.
