Research map / Time series / Long-term forecasting
Shallow RNN: Accurate Time-series Classification on Resource Constrained Devices
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 full-window MI-RNN with two-layer brick-based ShaRNN on top of it (MI-ShaRNN) | MI-RNN LSTM · Google-13 | accuracy (%): 93.16 -> 94.01 | Helped |
| replace MI-RNN with MI-ShaRNN | MI-RNN LSTM · DSA-19 | accuracy (%): 88.11 -> 87.36 | Hurt |
| replace MI-RNN with MI-ShaRNN | MI-RNN LSTM · HAR-6 | accuracy (%): 91.78 -> 94.02 | Helped |
| replace MI-RNN with MI-ShaRNN | MI-RNN LSTM · Google-13 | flops: 2.42M -> 0.59M | Helped |
| deploy MI-ShaRNN instead of MI-RNN on Cortex M4 | MI-RNN hidden 16 · Google-13 on Cortex M4 | inference time (ms): 226 -> 70.5 | Helped |
| deploy MI-ShaRNN instead of MI-RNN on Cortex M4 | MI-RNN hidden 16 · Google-13 on Cortex M4 | accuracy (%): 89.78 -> 91.42 | Helped |
| use first-order (M=2) instead of zeroth-order (M=1, truncation) approximation (ablation) | ShaRNN · Google-13 | approximation error / accuracy | Helped |
| use ShaRNN in both Listener and Speller instead of Listener only (ablation) | streaming LAS · TIMIT | phoneme error rate | Helped |
| reduce brick size k from 784 to 64 in ShaRNN LAS (ablation) | LAS · TIMIT | phoneme error rate: 0.22 -> 0.238 | Hurt |
| vary brick size k (ablation) | MI-ShaRNN · Google-13 | accuracy | No clear effect |
Closest papers by meaning
- Fast-Slow Recurrent Neural Networks (NeurIPS 2017 · similarity 0.89)
- Combining Recurrent, Convolutional, and Continuous-time Models with Linear State Space Layers (NeurIPS 2021 · similarity 0.89)
- Parallelizing Linear Recurrent Neural Nets Over Sequence Length (ICLR 2018 · similarity 0.88)
- Fully Neural Network Based Speech Recognition on Mobile and Embedded Devices (NeurIPS 2018 · similarity 0.88)
- Temporal Pyramid Recurrent Neural Network (AAAI 2020 · similarity 0.88)
- Accelerating Linear Recurrent Neural Networks for the Edge with Unstructured Sparsity (ICML 2025 · similarity 0.87)
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.