The current repository suggests a federated NILM disaggregator allowing to simulate distributed training to find a balance between high sampling rate required by deep models and the privacy of the consumers since no data is exchanged and the models are trained locally and aggregated at a central node.

The suggested code allows the simulation of federated training setups. The suggested extension is fully compatible with the new API of NILMtk. It allows specifying the different buildings contributing to the training as well as the number of randomly selected clients in each training round. The figure below illustrates the different steps followed by the suggested code. Furthermore, the FedAvg algorithm is used for aggregating the locally trained models. Thus, the suggested code assumes that the same data is available for different clients. The disaggregator is compatible with all seq2point models implemented in Deep-NILMtk, where the model type can be specified as a hyperparameter for the FL framework.

docker build -t experiment_setup .
docker run --gpus 'all' --name exp_exec experiment_setup
docker cp exp_exec:/home/guestuser/model_evaluation ./results

Bousbiat, H.; Himeur, Y.; Varlamis, I.; Bensaali, F.; Amira, A.; Neural Load Disaggregation:Meta-analysis, Federated Learning and Beyond. Energies 2023

Code released under the MIT Licence.