Research map / Architectures and efficiency / RNNs and normalization
Towards Hardware-Aware Tractable Learning of Probabilistic Models
Laura I Galindez Olascoaga, Wannes Meert, Nimish Shah, Marian Verhelst, Guy Van den Broeck
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 |
|---|---|---|---|
| scale precision down from 64-bit and prune features/sensors jointly (ablation) | hardware-aware PSDD/AC optimization · HAR, AC evaluation cost | computational cost | Improved |
| tolerate larger accuracy loss when choosing Pareto configuration (ablation) | hardware-aware PSDD/AC optimization · HAR, AC evaluation cost | computational cost: up to 3 orders of magnitude | Improved |
| prune features with AC simplification (Algorithm 2) (ablation) | hardware-aware PSDD/AC optimization · HAR | computational cost per feature pruned: 2 to 10 % | Improved |
| prune 11 of 28 features (ablation) | hardware-aware PSDD/AC optimization · HAR train | computational cost: 53% | Improved |
| use smaller model alpha3, prune 18 features, one sensor off, 32 bit, with 0.4% accuracy loss (ablation) | hardware-aware PSDD/AC optimization · HAR train | cost: 13% | Improved |
| evaluate Pareto configs on test data rather than train | hardware-aware PSDD/AC optimization · HAR | accuracy: within +-1% | No clear effect |
| random failure of feature sets (|F|/10, |F|/5, |F|/2) | hardware-aware PSDD/AC optimization · HAR Pareto configs | accuracy: -2% | Hurt |
| compare against TAN classifier | hardware-aware PSDD/AC optimization · HAR | cost and accuracy | Improved |
| switch operating points at run-time with activity-based policy | hardware-aware PSDD/AC optimization · HAR test | cost-accuracy trade-off | Improved |
| choose operating pt. 2 versus operating pt. 1 (full model) (ablation) | hardware-aware PSDD/AC optimization · Wilt | C_AC: 1 -> 0.07 | Improved |
| choose operating pt. 4 versus operating pt. 1 (ablation) | hardware-aware PSDD/AC optimization · Banknote | test accuracy: 95.6 -> 86.8 | Hurt |
| choose operating pt. 4 versus operating pt. 1 (ablation) | hardware-aware PSDD/AC optimization · Wilt | test accuracy: 97.5 -> 97.5 | No clear effect |
Closest papers by meaning
- On the Hardness of Approximating Distributions with Tractable Probabilistic Models (NeurIPS 2025 · similarity 0.85)
- Learning Tractable Probabilistic Models from Inconsistent Local Estimates (NeurIPS 2022 · similarity 0.84)
- TinyTrain: Resource-Aware Task-Adaptive Sparse Training of DNNs at the Data-Scarce Edge (ICML 2024 · similarity 0.83)
- Scaling Tractable Probabilistic Circuits: A Systems Perspective (ICML 2024 · similarity 0.82)
- QUTE: Quantifying Uncertainty in TinyML models with Early-exit-assisted ensembles for model-monitoring (ICML 2025 · similarity 0.82)
- The Expressivity Boundary of Probabilistic Circuits: A Comparison with Large Language Models (NeurIPS 2026 · similarity 0.82)
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.
