Efficient and Less Centralized Federated Learning

With the rapid growth in mobile computing, massive amounts of data and computing resources are now located at edge. To this end, Federated learning (FL) is becoming a widely adopted distributed machine (ML) paradigm, which aims to harness expanding skewed locally order develop rich informative models. In centralized FL, collection devices collaboratively solve ML task under coordination central server. However, existing FL frameworks make an over-simplistic assumption about network connectivity ignore communication bandwidth different links network. paper, we present study novel algorithm, mostly collaborate with other pairwise manner. Our nonparametric approach able exploit topology reduce bottlenecks. We evaluate our on various benchmarks demonstrate that method achieves 10\(\times \) better efficiency around 8% increase accuracy compared approach.