Tag
A matched-protocol comparison of synchronous ring gossip, FedAvg, local and centralized training for LSTM failure detectors on the NASA C-MAPSS turbofan benchmark, showing gossip as a practical serverless alternative to federated averaging that matches FedAvg performance without a coordinator.
CALM proposes a method for decentralized federated learning that uses class-wise agreement and label-gated disagreement modulation to handle non-IID data, improving teacher weighting and distillation trust without additional communication overhead.
This paper provides convergence theory for knowledge distillation in asynchronous peer-to-peer gossip learning networks, demonstrating that it contracts function disagreement and analyzing theoretical convergence rates.
The paper proposes CRAD, a class-wise reliability-aware distillation method for decentralized federated learning to handle heterogeneous architectures and non-IID data, achieving improved accuracy on image classification benchmarks.
This paper introduces RW-LoRA, a decentralized LoRA fine-tuning method using random walks to reduce communication and computation costs while achieving competitive performance on NLP tasks.
This paper proposes a unified framework for decentralized learning under communication constraints, introducing the DMFL-SQ algorithm that combines graph-based personalization, agnostic fairness, and compressed communication to achieve reduced communication while maintaining predictive performance and improving fairness across clients.
This paper studies decentralized multi-player Q-learning in episodic Markov decision processes under three forms of information asymmetry, proposing algorithms that achieve regret bounds matching the single-agent Q-learning rate up to logarithmic factors.
The paper introduces MEMOA, a decentralized strategy for massive online agents that achieves optimality via mean-field Nash equilibria, outperforming greedy baselines while scaling better than centralized approaches.