Dynamic Edge Association and Resource Allocation in Self-Organizing Hierarchical Federated Learning Networks

Federated Learning (FL) is a promising privacy-preserving distributed machine learning paradigm. However, communication inefficiency remains the key bottleneck that impedes its large-scale implementation. Recently, hierarchical FL (HFL) has been proposed in which data owners, i.e., workers, can first transmit their updated model parameters to edge servers for intermediate aggregation. This reduces instances of global and straggling workers. To enable efficient HFL, it important address issues association resource allocation context non-cooperative players, servers, owner. existing studies merely focus on static approaches do not consider dynamic interactions bounded rationalities players. In this paper, we propose game framework study dynamics self-organizing HFL networks. lower-level game, strategies workers are modelled using an evolutionary game. upper-level Stackelberg differential adopted owner decides optimal reward scheme given expected bandwidth control strategy server. Finally, provide numerical results validate our captures system under varying sources network heterogeneity.