In the paradigm of decentralized learning, a group of agents collaborate to learn a global model using a distributed dataset without a central server; nevertheless, it is severely challenged by the heterogeneity of the data distribution across the agents. For example, the data may be distributed non-independently and identically, and even be noised or poisoned. To address these data challenges, we propose ROSS, a novel robust decentralized stochastic learning algorithm based on Shapley values, in this paper. Specifically, in each round, each agent leverages its local gradient and the cross-gradients from its neighbors (i.e., the derivative of its local loss function and the derivatives of its neighbors’ local loss functions evaluated at its local model) to update its local model in a momentum-like manner, while we innovate in weighting the derivatives according to their contributions measured by Shapley values. We perform solid theoretical analysis to reveal the linear convergence speedup of our ROSS algorithm. We also verify the efficacy of our algorithm through extensive experiments on public datasets. Our results demonstrate that, in face of the above variety of data challenges, our ROSS algorithm have oblivious advantages over existing state-of-the-art proposals in terms of both convergence and prediction accuracy.

Diabetic Retinopathy (DR) represents a significant ophthalmic complication of diabetes, demanding accurate and interpretable diagnostic tools to mitigate vision loss through timely interventions. Existing supervised learning techniques, while successful in general image tasks, face critical limitations in medical contexts due to the high cost and complexity of detailed clinical annotations required for precise DR grading. In response, we propose DR-Reasoning, a novel multimodal vision-language reasoning framework tailored specifically for DR classification tasks. Our method selects the Qwen2.5-VL-7BInstruct model as the base model, and is trained on a real hospital dataset using Low-Rank Adaptation (LoRA) and Group Relative Policy Optimization (GRPO). DR-Reasoning integrates explicit reasoning chains into the diagnostic process, enabling effective capture of subtle pathological features in both fundus images and clinical descriptions. Extensive evaluations demonstrate that DRReasoning achieves state-of-the-art performance, substantially outperforming baseline methods across critical metrics such as AUC (92.79%) and specificity (97.69%).

In the paradigm of decentralized learning, a group of agents collaborates to learn a global model using distributed datasets without a central server. However, due to the heterogeneity of the local data across the different agents, learning a robust global model is rather challenging. Moreover, the collaboration of the agents relies on their gradient information exchange, which poses a risk of privacy leakage. In this paper, to address these issues, we propose PDSL, a novel privacy-preserved decentralized stochastic learning algorithm with heterogeneous data distribution. On one hand, we innovate in utilizing the notion of Shapley values such that each agent can precisely measure the contributions of its heterogeneous neighbors to the global learning goal; on the other hand, we leverage the notion of differential privacy to prevent each agent from suffering privacy leakage when it contributes gradient information to its neighbors. We conduct both solid theoretical analysis and extensive experiments to demonstrate the efficacy of our PDSL algorithm in terms of privacy preservation and convergence.