Code for the paper titled "FedNS: A Fast Sketching Newton-type Algorithm for Federated Learning"

This paper has been accepted at AAAI 2024. The code is based on FedNew, implemented by Matlab.

Run run_comparison.m to conduct experiments for the performance of related algorithms w.r.t. the communication rounds.
Run run_sketch_size.m to conduct experiments on the influence of the sketch size for FedNS and FedNDES.

Name		Name	Last commit message	Last commit date
Latest commit History 1 Commit
FedNDES.m		FedNDES.m
FedNS.m		FedNS.m
FedNewton.m		FedNewton.m
GD.m		GD.m
README.md		README.md
draw_image.m		draw_image.m
gen_sketch_mat.m		gen_sketch_mat.m
libsvmread.mexw64		libsvmread.mexw64
main_cleanDataset.m		main_cleanDataset.m
newton_ADMM.m		newton_ADMM.m
newton_ADMM_Hk.m		newton_ADMM_Hk.m
newton_ADMM_periodic_Hk.m		newton_ADMM_periodic_Hk.m
newton_QADMM.m		newton_QADMM.m
newton_QADMM_Hk.m		newton_QADMM_Hk.m
newton_QADMM_periodic_Hk.m		newton_QADMM_periodic_Hk.m
newton_zero.m		newton_zero.m
run_comparison.m		run_comparison.m
run_sketch_size.m		run_sketch_size.m
standard_newton.m		standard_newton.m
stochastic_quantization.m		stochastic_quantization.m

Provide feedback