Research map / Time series / Long-term forecasting
Dynamic Fractal Mamba: A Neural Renormalization Group Flow for Scale-Invariant Sequence Modeling
Shenglei Fang, Xianfang Sun, You Zhou
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 |
|---|---|---|---|
| add recursive weight-shared hierarchy over stacked Mamba (ablation) | DF-Mamba-TS · ETTm2 | MSE: 0.210 -> 0.158 | Improved |
| add weight sharing across recursion depths (ablation) | DF-Mamba-TS · ETTm2 | MSE: 0.180 -> 0.158 | Improved |
| use exponential instead of linear time-step scaling (ablation) | DF-Mamba-TS · ETTm2 | MSE: 0.194 -> 0.158 | Improved |
| use exponential instead of constant time-step scaling (ablation) | DF-Mamba-TS · ETTm2 | MSE: 0.186 -> 0.158 | Improved |
| use learned coarse-graining instead of naive pooling (ablation) | DF-Mamba-TS · ETTm2 | MSE: 0.169 -> 0.158 | Improved |
| use gated fusion instead of simple addition (ablation) | DF-Mamba-TS · ETTm2 | MSE: 0.163 -> 0.158 | Improved |
| add cross-scale consistency loss (ablation) | DF-Mamba-TS · ETTm2 | MSE: 0.167 -> 0.158 | Improved |
| add reconstruction loss (ablation) | DF-Mamba-TS · ETTm2 | MSE: 0.171 -> 0.158 | Improved |
| add both auxiliary losses to task loss (ablation) | DF-Mamba-TS · ETTm2 | MSE: 0.174 -> 0.158 | Improved |
| add consistency loss (ablation) | DF-Mamba · Breast histopathology (train L=256) | accuracy (%): 71.11 -> 77.78 | Improved |
| add exponential time-scaler (ablation) | DF-Mamba · Breast histopathology (train L=256) | accuracy (%): 71.11 -> 77.78 | Improved |
| add consistency loss (ablation) | DF-Mamba · Breast histopathology (train L=64) | accuracy (%): 66.67 -> 71.11 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Affirm — Affirm: Interactive Mamba with Adaptive Fourier Filters for Long-term Time Series Forecasting (AAAI 2025) · better on 2 of 4 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
- Routing Mamba: Scaling State Space Models with Mixture-of-Experts Projection (NeurIPS 2025 · similarity 0.86)
- Achilles' Heel of Mamba: Essential difficulties of the Mamba architecture demonstrated by synthetic data (NeurIPS 2025 · similarity 0.86)
- Parallel-in-Time Training of Recurrent Neural Networks for Dynamical Systems Reconstruction (NeurIPS 2026 · similarity 0.86)
- Block-Biased Mamba for Long-Range Sequence Processing (NeurIPS 2025 · similarity 0.85)
- LongMamba: Enhancing Mamba's Long-Context Capabilities via Training-Free Receptive Field Enlargement (ICLR 2025 · similarity 0.85)
- Mamba Can Learn Low-Dimensional Targets In-Context via Test-Time Feature Learning (NeurIPS 2026 · 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.
