advanced-math-trading/optimization-advanced

Multi-objective optimization balances competing goals—return, risk, and turnover—using techniques like NSGA-II and Pareto optimization. This Skill applies those methods to adaptive portfolio management, letting you explore trade-off surfaces and select strategies aligned with your constraints.

Yes. Distributional robust optimization (DRO) with CVaR protects against tail events by optimizing under uncertainty without assuming a fixed distribution. This Skill uses cvxpy and moment-DRO methods to build portfolios robust to worst-case scenarios within an ambiguity set.

Mixed-integer programming (MIP) encodes discrete decisions—position counts, buy/sell triggers, cardinality constraints—directly into the optimization model. This Skill integrates MIP via Gurobi to solve portfolio problems with hard combinatorial requirements that continuous methods cannot address.

Bayesian optimization iteratively samples the hyperparameter space, building a probabilistic model to identify promising regions with fewer evaluations than grid search. This Skill provides Bayesian tuning to adaptively find optimal strategy parameters, reducing computation while improving performance.

Yes. Online convex optimization (OGD, FTRL) updates portfolio weights incrementally as new data arrives, avoiding full recomputation. This Skill implements these methods for adaptive strategies that respond to live market streams without batch reprocessing.

Familiarity with convex optimization, Markowitz portfolio theory, and Black-Litterman models is helpful. This Skill links to those foundations and provides modular, load-on-demand components, so you can build up knowledge as you apply techniques to your trading problem.