A Hamiltonian, energy-conserving optimizer (ECDq=1) reduces initialization dependence and mean error compared to Adam for neural likelihood-ratio estimation in two collider-physics benchmarks.
Monte Carlo
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
hep-ph 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Optimizers for Stabilizing Likelihood-free Inference
A Hamiltonian, energy-conserving optimizer (ECDq=1) reduces initialization dependence and mean error compared to Adam for neural likelihood-ratio estimation in two collider-physics benchmarks.