Code and experiments for Algorithms and Learning for Fair Portfolio Design

Python implementation of algorithms and experiments described in Algorithms and Learning for Fair Portfolio Design by Emily Diana, Travis Dick, Hadi Elzayn, Michael Kearns, Aaron Roth, Zachary Schutzman, Saeed Sharifi-Malvajerdi, and Juba Ziani

Suggested citation: coming soon

---

Usage: python experiments_paper.py e1 e2 e3... where the e's are the experiments to run.

E.g. python experiments_paper.py 1 4 11 2 will run Experiments 1, 2, 4, and 11.

List of experiments and corresponding figures in the paper (if included)

• Experiment 1: No regret convergence time as a function of the step size.
• Experiment 2: Comparison of ILP, sparsify, greedy, greedy MM
• Experiment 3: Groups are drawn from gaussians with the same mean, different variance
• Experiment 4: Groups are drawn from gaussians with different means, same variance
• Experiment 5: Groups are drawn from gaussians, the second's mean and variance are scaled multiples of the first
• Experiment 6: One group is relatively small compared to the other
• Experiment 7: Generalization
• Experiment 8: Examining sparsification
• Experiment 9: Comparison of algorithms (Figure 2)
• Experiment 10: Product dependence (+timing)
• Experiment 11: Consumer dependence (+timing)
• Experiment 12: Generalization (Figure 3)

Edit configs.py to change important global variables, including the solver used by CVXPY. Currently supported: Gurobi and Mosek; other values will result in the default solvers being called.

Required packages are listed in requirements.txt and installed with pip install -r requirements.txt.