Applying bounded nonlinear functions (tanh, softsign, arctan) to DNN weights makes models tolerate random bit flips at BER 1e-5 with only a few points of accuracy loss.
Proact: Progressive training for hybrid clipped activation function to enhance resilience of dnns,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Enhancing Neural Network Robustness Against Fault Injection Through Non-linear Weight Transformations
Applying bounded nonlinear functions (tanh, softsign, arctan) to DNN weights makes models tolerate random bit flips at BER 1e-5 with only a few points of accuracy loss.