EA-HAS-Bench's SMI-based per-epoch energy measurements correlate poorly with an external power meter (Pearson 0.64) due to under-sampling in 39% of epochs, and the paper proposes design principles and a calibration method for more reliable energy-aware benchmarking.
ImageNet classification with deep convolutional neural networks,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Guidelines for the Quality Assessment of Energy-Aware NAS Benchmarks
EA-HAS-Bench's SMI-based per-epoch energy measurements correlate poorly with an external power meter (Pearson 0.64) due to under-sampling in 39% of epochs, and the paper proposes design principles and a calibration method for more reliable energy-aware benchmarking.