Federated Learning Library: https://fedml.ai
machine-learning
privacy
computer-vision
semi-supervised-learning
transfer-learning
wireless-communication
distributed-optimization
interpretability
neural-architecture-search
federated-learning
continual-learning
vertical-federated-learning
non-iid
decentralized-federated-learning
hierarchical-federated-learning
adversarial-attack-and-defense
communication-efficiency
straggler-problem
computation-efficiency
incentive-mechanism
-
Updated
Jul 20, 2021
For the moment, the nc_benchmark generator function allows for a nc_first_task option, which is good for pre-training in the class-incremental learning scenario. However, the same kind of option is not available if one wants to pretrain in the task-incremental scenario. It would be nice to have an option that could be used together with task_labels=True and allows for pretraining on multiple tasks