Research map / Probabilistic and scientific modeling / Dynamical systems
Fault Diagnosis of Irregular Sequences by Adjoint Learning in Continuous-Time Model Space
Xiren Zhou, Chuyang Wei, Ao Chen, Shikang Liu, Xiangyu Wang, Huanhuan Chen
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 adjoint optimization (fitting + discrimination losses, learnable metric) to CT-Res (ablation) | CT-Res · CWRU-D (50% missing) | accuracy (%): 87.05 -> 96.61 | Improved |
| add adjoint optimization to CT-Res (ablation) | CT-Res · TBV (50% missing) | accuracy (%): 64.59 -> 78.76 | Improved |
| add adjoint optimization to CT-Res (ablation) | CT-Res · SU-G20 (50% missing) | accuracy (%): 92.33 -> 96.22 | Improved |
| use continuous-time ODE-embedded reservoir (CT-Res) instead of discrete-time adjoint ESN for fitting (ablation) | proposed approach · CWRU-D (50% missing) | accuracy (%): 78.08 -> 96.61 | Improved |
| use continuous-time ODE-embedded reservoir (CT-Res) instead of discrete-time adjoint ESN (ablation) | proposed approach · TBV (50% missing) | accuracy (%): 62.29 -> 78.76 | Improved |
| optimize with adjoint ESN instead of direct backpropagation through ODE solver (ablation) | CT-Res · CWRU training time | training time (s): 1954.37 -> 10.24 | Improved |
| optimize with adjoint ESN instead of direct backpropagation (ablation) | CT-Res · WHU training time | training time (s): 686.28 -> 3.72 | Improved |
| increase missing rate from 10% to 70% (ablation) | proposed approach · TBV | accuracy (%): 95.58 -> 51.29 | Hurt |
| increase missing rate from 10% to 70% (ablation) | proposed approach · CWRU average | accuracy (%): 96.92 -> 81.34 | Hurt |
| proposed approach vs Warpformer on SU-B30 | proposed approach · SU-B30 (50% missing) | accuracy (%): 99.14 -> 99.25 | No clear effect |
| proposed approach vs GRU-∆t on SU-G20 | proposed approach · SU-G20 (50% missing) | accuracy (%): 93.39 -> 96.22 | Improved |
| use ODE-RNN as baseline on SU-G20 versus proposed | ODE-RNN · SU-G20 (50% missing) | accuracy (%): 96.22 -> 62.12 | Hurt |
Closest papers by meaning
- Efficient Anomaly Detection of Irregular Sequences in Ct-Echo Model Space (AAAI 2025 · similarity 0.86)
- Faithful Embeddings of Irregular and Asynchronous Data for Online Log-NCDEs (NeurIPS 2026 · similarity 0.82)
- MD tree: a model-diagnostic tree grown on loss landscape (ICML 2024 · similarity 0.82)
- Sequence Modeling with Spectral Mean Flows (NeurIPS 2025 · similarity 0.82)
- Efficient Training of Neural Fractional-Order Differential Equation via Adjoint Backpropagation (AAAI 2025 · similarity 0.81)
- Parallel-in-Time Training of Recurrent Neural Networks for Dynamical Systems Reconstruction (NeurIPS 2026 · similarity 0.81)
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.
