Research map / Time series / Long-term forecasting
Anti-Aliasing Matters: A Dynamic Network for Time Series Forecasting
Heng Zhou, Xin Sun, Chao Li
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 |
|---|---|---|---|
| replace depthwise+pointwise conv downsampling with linear layers (ablation) | DMANet · Unemp | MSE: 0.064 -> 0.081 | Hurt |
| use standard strided convolution instead of depthwise+pointwise (ablation) | DMANet · Unemp | MSE: 0.064 -> 0.075 | Hurt |
| use frequency-domain upsampling with channel expansion before zero-padding (ablation) | DMANet · Unemp | MSE: 0.077 (Linear Up) -> 0.064 | Improved |
| add RevIN normalization (ablation) | DMANet · NASDAQ | MSE: 2.337 -> 0.177 | Improved |
| use frequency-domain MAE loss instead of MSE loss (ablation) | DMANet · NASDAQ | MSE: 0.195 -> 0.177 | Improved |
| add pre-sampling anti-aliasing filter (ablation) | DMANet generic baseline · Unemp | MSE: 0.075 -> 0.064 | Improved |
| add post-sampling frequency interpolation (ablation) | DMANet generic baseline · Unemp | MSE: 0.069 -> 0.064 | Improved |
| use ESR-based dynamic low-pass mask instead of Max/Random/classical filters (ablation) | DMANet · PEMS08 | MSE: 0.117 (Max) -> 0.090 | Improved |
| plug ESR-based anti-aliasing filter into TimeMixer and MICN | TimeMixer, MICN · multiple datasets | MSE/MAE | Improved |
| add anti-aliasing filter on a high-frequency injected ETTh1 signal | DMANet · ETTh1 spectral injection | High-Frequency Capture: 99.8% HFC vs 76.9% FilterNet | Improved |
| anti-aliasing filter overhead | DMANet · synthetic data | latency: worst-case increase of only 2.4% | No clear effect |
| DMANet vs iTransformer on high-dimensional Solar-Energy | DMANet · Solar-Energy (L=96) | MSE: 0.227 vs TimeMixer 0.216 | Mixed |
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.
- TimeMixer — TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting (ICLR 2024) · better on 6 of 7 shared results
- TVNet — TVNet: A Novel Time Series Analysis Method Based on Dynamic Convolution and 3D-Variation (ICLR 2025) · better on 2 of 2 shared results
- iTransformer — iTransformer: Inverted Transformers Are Effective for Time Series Forecasting (ICLR 2024) · better on 2 of 2 shared results
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
- FilterNet: Harnessing Frequency Filters for Time Series Forecasting (NeurIPS 2024 · similarity 0.86)
- TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting (ICLR 2024 · similarity 0.86)
- TimeCHEAT: A Channel Harmony Strategy for Irregularly Sampled Multivariate Time Series Analysis (AAAI 2025 · similarity 0.86)
- TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting (NeurIPS 2025 · similarity 0.86)
- TimeMixer++: A General Time Series Pattern Machine for Universal Predictive Analysis (ICLR 2025 · similarity 0.85)
- Apollo-Forecast: Overcoming Aliasing and Inference Speed Challenges in Language Models for Time Series Forecasting (AAAI 2025 · similarity 0.85)
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.
