This is the authors' abstract. We don't add key points for this paper.
Split learning enables privacy-preserving collaborative training of crop disease classifiers by partitioning a neural network between resource-constrained rural devices and a remote aggregation server, yet its feasibility under intermittent connectivity remains unquantified. The analysis characterises the operating characteristics of the design under a declared variance structure (between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.640), with power estimated across 2000 design cells. Results indicate that the intraclass correlation of 0.64 substantially inflates the cluster count required to detect small accuracy differences, with 54 clusters needed for 5 percentage-point effects at 80% power. The findings establish boundary conditions under which split learning is communication-viable for rural plant disease diagnosis and identify the split-layer position that minimises uplink cost without degrading accuracy beyond a specified tolerance.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Ceesay-Jallow et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: