Research map / Time series / Causal and generative time series
DeepSITH: Efficient Learning via Decomposition of What and When Across Time Scales
Brandon G Jacques, Zoran Tiganj, Marc Howard, Per B Sederberg
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 |
|---|---|---|---|
| use DeepSITH instead of LSTM | recurrent baseline · psMNIST | accuracy: 90.20 -> 97.36 | Improved |
| use DeepSITH instead of LMU | recurrent baseline · psMNIST | accuracy: 97.15 -> 97.36 | No clear effect |
| use DeepSITH instead of coRNN | recurrent baseline · psMNIST | accuracy: 97.34 -> 97.36 | No clear effect |
| use DeepSITH instead of coRNN | recurrent baseline · sMNIST | accuracy: 99.40 -> 99.32 | No clear effect |
| use DeepSITH instead of LSTM | recurrent baseline · sMNIST | accuracy: 98.90 -> 99.32 | Improved |
| use DeepSITH instead of LMU, LSTM, coRNN | recurrent baseline · Adding Problem | MSE learning speed | Improved |
| increase signal length T from 100 to 5000 | LMU and LSTM · Adding Problem | MSE | Hurt |
| use DeepSITH instead of LMU, LSTM, coRNN | recurrent baseline · Mackey-Glass | NRMSE | Improved |
| increase tau/prediction distance from 17/15 to 85/75 | DeepSITH · Mackey-Glass | NRMSE | No clear effect |
| increase noise duration | LSTM · Hateful-8 | accuracy | Hurt |
| increase noise duration | DeepSITH · Hateful-8 | accuracy | No clear effect |
| use DeepSITH or LMU instead of coRNN | recurrent baseline · Hateful-8 | training speed | Improved |
Closest papers by meaning
- Combining Recurrent, Convolutional, and Continuous-time Models with Linear State Space Layers (NeurIPS 2021 · similarity 0.86)
- A deep convolutional neural network that is invariant to time rescaling (ICML 2022 · similarity 0.86)
- Weight-Space Linear Recurrent Neural Networks (ICLR 2026 · similarity 0.86)
- Emergent mechanisms for long timescales depend on training curriculum and affect performance in memory tasks (ICLR 2024 · similarity 0.85)
- Deep Reinforcement Learning with Time-Scale Invariant Memory (AAAI 2025 · similarity 0.84)
- Learning dynamics in linear recurrent neural networks (ICML 2025 · similarity 0.84)
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.
