This repo is one of the projects in the 15-780 course (Graduate Artificial Intelligence, Spring'22) of CMU.
It replicates a part of Figure 2 of the paper Can We Predict the Election Outcome from Sampled Votes? in AAAI-20.
It also computes the top-k average-case predictability. (Can we predict the top-k election outcome instead of the full outcome?) In some cases, for example, we only care about the top-1 election outcome.
To run
Simply compile the single file voting_top_k.cpp and run to get the result. It takes about 30 seconds.
To plot the result, copy the output of voting_top_k.cpp to the corresponding part of plot.py, and run plot.py.
Extension
voting_top_k.cpp can be easily extended for other voting rules and other parameters (e.g., changing the parameters n=1000, m=5, #draws=10^6, or the plotting number of samples in the paper).