Research map / Learning theory and algorithms / Prediction and decision losses
Subtask Gated Networks for Non-Intrusive Load Monitoring
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 |
|---|---|---|---|
| Add an on/off classification subtask and use its probability to gate regression output | Seq2Seq for NILM · REDD | MAE: 17.34 % | Helped |
| Add an on/off classification subtask and use its probability to gate regression output | Seq2Seq for NILM · REDD | SAEδ: 28.34 % | Helped |
| Add an on/off classification subtask and use its probability to gate regression output | Seq2Seq for NILM · UK-DALE | MAE: 21.72 % | Helped |
| Add an on/off classification subtask and use its probability to gate regression output | Seq2Seq for NILM · UK-DALE | SAEδ: 18.09 % | Helped |
| Add learnable standby power (ablation) | SGN · REDD | MAE: 19.09 -> 18.86 | Helped |
| Add learnable standby power (ablation) | SGN · REDD | SAEδ: 11.97 -> 12.23 | Hurt |
| Add learnable standby power (ablation) | SGN · UK-DALE | MAE: 11.39 -> 10.89 | Helped |
| Add learnable standby power (ablation) | SGN · UK-DALE | SAEδ: 7.66 -> 7.71 | Hurt |
| Replace soft gating with thresholded hard gating (ablation) | SGN · REDD | MAE: 19.09 -> 20.33 | Hurt |
| Replace soft gating with thresholded hard gating (ablation) | SGN · UK-DALE | SAEδ: 7.66 -> 7.55 | Helped |
| Remove the on/off classification loss (ablation) | SGN and its variants · REDD | MAE: SGN: 19.09 -> 24.09; SGN-sp: 18.86 -> 24.99; Hard SGN: 20.33 -> 28.16; Hard SGN-sp: 21.56 -> 33.55 | Hurt |
| Remove the on/off classification loss (ablation) | SGN and its variants · UK-DALE | MAE: SGN: 11.39 -> 17.17; SGN-sp: 10.89 -> 12.64; Hard SGN: 11.00 -> 20.12; Hard SGN-sp: 13.21 -> 30.25 | Hurt |
Closest papers by meaning
- Sequence-to-Point Learning With Neural Networks for Non-Intrusive Load Monitoring (AAAI 2018 · similarity 0.92)
- MSDC: Exploiting Multi-State Power Consumption in Non-intrusive Load Monitoring Based on a Dual-CNN Model (AAAI 2023 · similarity 0.89)
- Deep Latent Generative Models for Energy Disaggregation (AAAI 2019 · similarity 0.85)
- Model-Free Iterative Temporal Appliance Discovery for Unsupervised Electricity Disaggregation (AAAI 2018 · similarity 0.84)
- PowerPM: Foundation Model for Power Systems (NeurIPS 2024 · similarity 0.79)
- PowerGraph: A power grid benchmark dataset for graph neural networks (NeurIPS 2024 · similarity 0.78)
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.