Codes for "Distributed Experimental Design Networks"

Y Li, L Su, C Joe-Wong, E Yeh, and S Ioannidis, "Distributed Experimental Design Networks", IEEE INFOCOM 2024

Please cite this paper if you intend to use this code for your research.

• Dependencies
• Usage

Dependencies

The dependencies are specified in the requirements.txt file.

Usage

ProbGenerate.py is to initialize the problems over different networks/topologies. Some execution examples are shown in run_Prob.

After initialize the problem, we use distributedSolver.py to solve it through both centralized and distributed algorithm. Specifically, it implements MaxTP, MaxFair, FW, PGA and their corresponding distributed versions: DMaxTP, DMaxFair, DFW, DPGA introduced by our paper. Some execution examples are shown in run_Solver.

After solving the problem, we could use Beta.py to calculate the average norm of estimation error, where the estimation error is the difference between the true model and the MAP estimator. Some execution examples are shown in run_beta.

plot_utility.py is to plot bar figures in the paper.

plot_stepsize.py is to plot stepsize effect in the paper. plot_sens.py is to plot other sensitivity in the paper. Some execution examples are shown in run_sens.