Module dro 

Distributionally Robust Optimization (DRO).

This module provides optimization algorithms that are robust against distributional uncertainty. Rather than assuming that the true data distribution is exactly the empirical distribution P_N, DRO seeks a decision that minimises the worst-case expected loss over an ambiguity set of plausible distributions.

Sub-modules

• types: shared configuration and result types
• wasserstein_dro: Wasserstein-ball DRO (Esfahani & Kuhn 2018)
• cvar_dro: CVaR-based DRO (Rockafellar & Uryasev 2000)

Quick start

use scirs2_optimize::dro::{portfolio_dro, solve_cvar_dro, DroConfig};

// Synthetic return data (3 assets, 20 observations).
let returns: Vec<Vec<f64>> = (0..20)
    .map(|i| vec![0.01 * i as f64, 0.02, -0.005 * i as f64])
    .collect();

// Wasserstein DRO portfolio (ε = 0.05).
let result = portfolio_dro(&returns, 0.05, None).expect("dro ok");
println!("Robust weights: {:?}", result.optimal_weights);
println!("Worst-case loss: {:.4}", result.worst_case_loss);

// CVaR-DRO with α = 0.9.
let samples: Vec<Vec<f64>> = (0..20).map(|i| vec![0.01 * i as f64, 0.02]).collect();
let cvar_result = solve_cvar_dro(2, &samples, 0.9, 0.1, None).expect("cvar dro ok");
println!("CVaR-DRO weights: {:?}", cvar_result.optimal_weights);

References

• Esfahani, P. M. & Kuhn, D. (2018). “Data-driven distributionally robust optimization using the Wasserstein metric.” Mathematical Programming.
• Rockafellar, R. T. & Uryasev, S. (2000). “Optimization of conditional value-at-risk.” Journal of Risk.

Re-exports

pub use cvar_dro::solve_cvar_dro;

pub use cvar_dro::CvarDro;

pub use cvar_dro::CvarEstimator;

pub use types::DroConfig;

pub use types::DroResult;

pub use types::DroSolver;

pub use types::RobustObjective;

pub use types::WassersteinBall;

pub use wasserstein_dro::portfolio_dro;

pub use wasserstein_dro::portfolio_erm;

pub use wasserstein_dro::WassersteinDro;

Modules

cvar_dro

CVaR-based Distributionally Robust Optimization.

types

Types for Distributionally Robust Optimization (DRO).

wasserstein_dro

Wasserstein Distributionally Robust Optimization.