Research map / Probabilistic and scientific modeling / Causal discovery
Frequency-Domain Better than Time-Domain for Causal Structure Recovery in Dynamical Systems on Networks
Mohammed Tuhin Rana, Mishfad Shaikh Veedu, James Melbourne, Murti Salapaka
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 |
|---|---|---|---|
| estimate Wiener filters in frequency domain (FFT) instead of time domain | W-PC · 25 random networks (large samples) | computation time: around an order of magnitude lower | Improved |
| estimate Wiener filters in frequency domain instead of time domain | W-PC · 25 random 6-node DAGs | average error rate | Improved |
| use frequency-domain estimation instead of time-domain | W-PC · 25 random 6-node DAGs | samples needed | Improved |
| use Wiener phase information in frequency domain | Wiener Phase vs TD W-PC · 25 random 6-node DAGs | average error rate | Improved |
| use FFT-WPC instead of TPC | time-series causal discovery · 20 nodes synthetic | F1 (%) and run time (s): 78.57 -> 100; 4016.57 -> 1011.62 | Improved |
| use Wiener Phase instead of TPC | time-series causal discovery · 20 nodes synthetic | run time (s): 4016.57 -> 1.37 | Improved |
| use FFT-WPC instead of TPC | time-series causal discovery · River-runoff (real) | run time (s): 1548.29 -> 9.041 | Improved |
| use Wiener Phase instead of FFT-WPC | time-series causal discovery · River-runoff (real) | F1 (%): 75 -> 48.3 | Hurt |
| use Wiener Phase instead of GC | time-series causal discovery · River-runoff (real) | F1 (%): 33 -> 48.3 | Improved |
| use Wiener Phase instead of FFT-WPC | time-series causal discovery · 50 nodes synthetic network | run time (s): 142652 -> 19.49 | Improved |
| violate the phase alignment assumption (Wiener Phase) (ablation) | Wiener Phase · 20 nodes, 128000 samples | combined score CS (%): 92.69 | No clear effect |
| use Wiener Phase and FFT-WPC instead of GC when present value depends on present value | time-series causal discovery · 6-node zero-lag network | combined score CS (%): 80 -> 100 | Improved |
Closest papers by meaning
- Multi-domain Causal Structure Learning in Linear Systems (NeurIPS 2018 · similarity 0.89)
- CausalDynamics: A large‐scale benchmark for structural discovery of dynamical causal models (NeurIPS 2025 · similarity 0.88)
- Learning the Causal Structure of Networked Dynamical Systems under Latent Nodes and Structured Noise (AAAI 2024 · similarity 0.87)
- Discovering Mixtures of Structural Causal Models from Time Series Data (ICML 2024 · similarity 0.86)
- Learning Causal Relations from Subsampled Time Series with Two Time-Slices (ICML 2024 · similarity 0.86)
- Generating Hypotheses of Dynamic Causal Graphs in Neuroscience: Leveraging Generative Factor Models of Observed Time Series (ICML 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.
