Given only prediction access, an adversary can exactly recover the weights of a two-layer ReLU network, and semi-supervised learning makes accuracy extraction far more query-efficient.
The unreasonable effectiveness of data,
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
2019 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
High Accuracy and High Fidelity Extraction of Neural Networks
Given only prediction access, an adversary can exactly recover the weights of a two-layer ReLU network, and semi-supervised learning makes accuracy extraction far more query-efficient.