Research map / Architectures and efficiency / Neural architecture search
Hardware-adaptive Efficient Latency Prediction for NAS via Meta-Learning
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 |
|---|---|---|---|
| Adapt the latency predictor with 10 target-device reference measurements instead of training BRP-NAS with 900 architecture-latency pairs | MetaD2A latency-constrained NAS · CIFAR-100, NAS-Bench-201, unseen Eyeriss and FPGA; unseen Xeon Gold 6226 | closeness to latency constraint: 10 vs 900 samples | Mixed |
| Use HELP rather than BRP-NAS as the latency predictor | MetaD2A · CIFAR-100, NAS-Bench-201, Eyeriss, constraint 7 ms | latency (ms): 9.1 -> 5.5 | Helped |
| Use HELP rather than BRP-NAS as the latency predictor | MetaD2A · CIFAR-100, NAS-Bench-201, FPGA, constraint 5 ms | latency (ms): 7.2 -> 4.7 | Helped |
| Use HELP rather than BRP-NAS as the latency predictor | MetaD2A · CIFAR-100, NAS-Bench-201, FPGA, constraint 6 ms | latency (ms): 7.4 -> 5.9 | Helped |
| Use HELP rather than BRP-NAS as the latency predictor | MetaD2A · CIFAR-100, NAS-Bench-201, Xeon Gold 6226, constraint 8 ms | latency (ms): 9.5 -> 7.7 | Helped |
| Use HELP rather than BRP-NAS as the latency predictor | MetaD2A · CIFAR-100, NAS-Bench-201, Xeon Gold 6226, constraint 11 ms | latency (ms): 8.7 -> 11.0 | Hurt |
| Use HELP rather than BRP-NAS as the latency predictor | MetaD2A · CIFAR-100, NAS-Bench-201, Xeon Gold 6226, constraint 14 ms | latency (ms): 17.0 -> 13.9 | Helped |
| Combine HELP with OFA for latency prediction on an unseen platform | OFA · ImageNet, MobileNetV3, Jetson AGX Xavier | total NAS cost (wall clock): 1.3M$\times$ faster than MnasNet-A1 | Helped |
Closest papers by meaning
- BRP-NAS: Prediction-based NAS using GCNs (NeurIPS 2020 · similarity 0.90)
- HW-NAS-Bench: Hardware-Aware Neural Architecture Search Benchmark (ICLR 2021 · similarity 0.89)
- CloserToMe: A Unified Framework for Accurate and Transferable Latency Prediction Across Heterogeneous Devices (AAAI 2026 · similarity 0.87)
- Rapid Model Architecture Adaption for Meta-Learning (NeurIPS 2022 · similarity 0.86)
- NAS-Bench-201: Extending the Scope of Reproducible Neural Architecture Search (ICLR 2020 · similarity 0.86)
- FP-NAS: Fast Probabilistic Neural Architecture Search (CVPR 2021 · similarity 0.86)
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.