Sensitivity-ranked remapping of kernels in resistive crossbars limits line-resistance accuracy drop from 5.6% to 2.1-2.9% on VGG16/CIFAR10 without retraining.
Training and operation of an integrated neuromorphic network based on metal-oxide memristors
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.ET 1years
2019 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
X-CHANGR: Changing Memristive Crossbar Mapping for Mitigating Line-Resistance Induced Accuracy Degradation in Deep Neural Networks
Sensitivity-ranked remapping of kernels in resistive crossbars limits line-resistance accuracy drop from 5.6% to 2.1-2.9% on VGG16/CIFAR10 without retraining.