This repository accompanies our paper, Regulating Accuracy-Efficiency Trade-Offs in Distributed Machine Learning Systems, presented at ICML 2020's Law and Machine Learning Workshop.
We also include the simple scratch code we used to generate the graph figures in the paper. In order to run this code, please be sure to have Julialang >=1.4 installed (follow the instructions on their website) and follow these steps:
Open a terminal / command line program of your choice. Clone this directory to a convenient location, and then change directories into the
We need to then make sure that the appropriate
juliadependencies are loaded on your machine. To do so, run
julia setup.jl. Note: This only needs to be done the first time you run the graphing code.
To run the code we used for generating Figure 3, which explains sampling via flipping a biased coin, run
julia coin.jl. This will output the resulting figure in
flip.pdfincluded in this repository is the one we use in the paper.