Deep-Relative-Trust-Based Diffusion for Decentralized Deep Learning
This addresses the challenge of efficient decentralized learning for agents with local data, offering a novel approach that enhances performance in scenarios like image classification.
The paper tackled the problem of decentralized deep learning by proposing a new algorithm, DRT diffusion, that encourages consensus of neural network outputs instead of parameters, which improved generalization in image classification tasks, especially with sparse topologies.
Decentralized learning strategies allow a collection of agents to learn efficiently from local data sets without the need for central aggregation or orchestration. Current decentralized learning paradigms typically rely on an averaging mechanism to encourage agreement in the parameter space. We argue that in the context of deep neural networks, which are often over-parameterized, encouraging consensus of the neural network outputs, as opposed to their parameters can be more appropriate. This motivates the development of a new decentralized learning algorithm, termed DRT diffusion, based on deep relative trust (DRT), a recently introduced similarity measure for neural networks. We provide convergence analysis for the proposed strategy, and numerically establish its benefit to generalization, especially with sparse topologies, in an image classification task.