Improving $(\alpha, f)$-Byzantine Resilience in Federated Learning via layerwise aggregation and cosine distance

Open in new window