For a simulated Allora-style network, grid search finds optimal slopes p=3 (regression) and p=5 (classification), EMA alpha=0.1, and reward slopes pi=pf=3, pr=1.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Optimizing Decentralized Online Learning for Supervised Regression and Classification Problems
For a simulated Allora-style network, grid search finds optimal slopes p=3 (regression) and p=5 (classification), EMA alpha=0.1, and reward slopes pi=pf=3, pr=1.