The microcanonical entropy of the XOR network's loss landscape peaks at discrete loss values, and these entropic barriers vanish when more hidden neurons are added.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cond-mat.dis-nn 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Dissecting a Small Artificial Neural Network
The microcanonical entropy of the XOR network's loss landscape peaks at discrete loss values, and these entropic barriers vanish when more hidden neurons are added.