Research map / Time series / Causal and generative time series
Multiple Instance Learning for Efficient Sequential Data Classification on Resource-constrained Devices
Don Dennis, Chirag Pabbaraju, Harsha Vardhan Simhadri, Prateek Jain
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 |
|---|---|---|---|
| train on MIL instance windows (MI-RNN) instead of full sequence (ablation) | LSTM · GesturePod-6 (hidden 32) | accuracy: 94.04 -> 98.27 | Improved |
| train on MIL instance windows (MI-RNN) (ablation) | LSTM · Google-13 (hidden 16) | accuracy: 86.99 -> 89.78 | Improved |
| MI-RNN on larger hidden size (ablation) | LSTM · HAR-6 (hidden 16) | accuracy: 92.90 -> 93.89 | Improved |
| iteratively refine instance labels (MI-RNN vs round 0) (ablation) | MI-RNN · HAR-6 (hidden 8) | accuracy: 90.83 -> 91.92 | Improved |
| iteratively refine instance labels (MI-RNN vs round 0) (ablation) | MI-RNN · DSA-19 (hidden 32) | accuracy: 86.97 -> 87.01 | No clear effect |
| iterative refinement of instance labels (ablation) | MI-RNN · Google-13 (hidden 64) | accuracy: 91.80 -> 92.61 (hidden 32) | Improved |
| add early prediction (E-RNN) without MIL (ablation) | LSTM · GesturePod-6 (hidden 32) | accuracy: 94.04 -> 84.48 | Hurt |
| train MIL and early classification jointly (EMI-RNN) instead of separately (ablation) | MI-RNN + E-RNN · GesturePod-6 | accuracy | Improved |
| use EMI-RNN instead of LSTM, smaller hidden size (ablation) | LSTM · GesturePod-6 | inference speedup: 72x less computation; 97.13 -> 98.00 | Improved |
| use EMI-RNN instead of LSTM (ablation) | LSTM · Google-13, Raspberry Pi 0 (hidden 32) | prediction time (ms): 74.46 -> 14.96 | Improved |
Closest papers by meaning
- TinyFoA: Memory Efficient Forward-Only Algorithm for On-Device Learning (AAAI 2025 · similarity 0.83)
- TinyTrain: Resource-Aware Task-Adaptive Sparse Training of DNNs at the Data-Scarce Edge (ICML 2024 · similarity 0.82)
- TAIL-MIL: Time-Aware and Instance-Learnable Multiple Instance Learning for Multivariate Time Series Anomaly Detection (AAAI 2025 · similarity 0.82)
- TimeMIL: Advancing Multivariate Time Series Classification via a Time-aware Multiple Instance Learning (ICML 2024 · similarity 0.82)
- Transformers are almost optimal metalearners for linear classification (NeurIPS 2025 · similarity 0.81)
- Energy Efficient Streaming Time Series Classification with Attentive Power Iteration (AAAI 2024 · similarity 0.80)
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.
