FedLoDrop: Federated LoRA with Dropout for Generalized LLM Fine-tuning

approximation of one-to-one matching that leverages Sinkhorn iterations on a