A Hopfield-type spin glass constructed from a feedforward network yields replica overlap statistics that serve as a per-instance descriptor of the network's generalization, capacity, and robustness.
Quantum langevin dynamics for optimization
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cond-mat.dis-nn 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A Spin Glass Characterization of Neural Networks
A Hopfield-type spin glass constructed from a feedforward network yields replica overlap statistics that serve as a per-instance descriptor of the network's generalization, capacity, and robustness.