Federated Learning and Distillation Benchmarks for 3D Point Clouds

A new benchmark evaluates 13 federated learning algorithms and 10 knowledge distillation objectives for 3D point cloud classification. Results show that while non-IID data skews federated learning performance, combining it with distillation can produce student models 74.51% smaller and twice as fast.