CrossNAS combines single-path one-shot weight sharing with evolutionary search to co-optimize neural architectures, mixed-precision quantization, and analog PIM circuit parameters, achieving higher accuracy and lower energy-delay product than prior PIM-NAS methods on CIFAR-10 and CIFAR-100.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.ET 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
CrossNAS: A Cross-Layer Neural Architecture Search Framework for PIM Systems
CrossNAS combines single-path one-shot weight sharing with evolutionary search to co-optimize neural architectures, mixed-precision quantization, and analog PIM circuit parameters, achieving higher accuracy and lower energy-delay product than prior PIM-NAS methods on CIFAR-10 and CIFAR-100.