Memory-Shaped Optimization

Cross-source consensus on Memory-Shaped Optimization from 1 sources and 6 claims.

1 sources · 6 claims

How it works

SHAPE's shaped potential includes a memory-induced term, an anchor term, and an optional barrier or exclusion term. — When Descent Is Too Stable: Event-Triggered Hamiltonian Learning to Optimize
In the multi-well study, removing memory reduced success rate and worsened best gap. — When Descent Is Too Stable: Event-Triggered Hamiltonian Learning to Optimize
SHAPE uses memory to summarize previously visited regions at each stage. — When Descent Is Too Stable: Event-Triggered Hamiltonian Learning to Optimize
Scalable memory updates remain an open problem for high-dimensional spaces. — When Descent Is Too Stable: Event-Triggered Hamiltonian Learning to Optimize
SHAPE is hypothesized to use memory-shaped energy to avoid repeatedly refining uninformative basins. — When Descent Is Too Stable: Event-Triggered Hamiltonian Learning to Optimize
The empirical pattern indicates that memory improves success and best gap on multi-well and basin-escape tasks. — When Descent Is Too Stable: Event-Triggered Hamiltonian Learning to Optimize