Code and derived results for a federated irrigation scheduler with an accounted differential privacy budget

Description

Analysis code and derived result files for the study "Federated irrigation scheduling for smallholder farms with an accounted differential privacy budget". The archive contains the Python pipeline that produces every number reported in Section 4 of the manuscript: federated training with per-record gradient clipping and calibrated Gaussian noise, a Rényi differential privacy accountant, seven baselines, an eight-configuration ablation, a comparison of three aggregation rules, intermittent-participation and uplink-quantisation studies, an FAO-56 soil-water balance, and external validation on measured field data. One CSV per reported table is included, together with the six figures as rendered.

Categories

Licence