Privacy-preserving federated learning across edge sites. Energy forecasting · Anomaly detection · FedAvg aggregation.
Load sample data or upload your own CSV files
Predicted vs actual consumption — select a site
Samples exceeding 2.5σ residual threshold per site
Only model weights are shared — raw data never leaves the edge site
Loss convergence · 64→32 neural network
Raw energy data never leaves each site. Only neural network weight tensors are shared with the global server.
Two CSV files required to run the pipeline
Drag & drop both CSV files here
or click to browse
Loading pipeline data…