Chisme uses cosine similarity between model updates to weight how much a gossip-learning client trusts incoming models, improving convergence and fairness under heterogeneous data.
Like attracts like: Personalized federated learning in decentralized edge computing,
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
-
Chisme: Heterogeneity-Aware Gossip Learning
Chisme uses cosine similarity between model updates to weight how much a gossip-learning client trusts incoming models, improving convergence and fairness under heterogeneous data.