{"paper_id":"2411.12323","title":"Mirror Descent Algorithms for Risk Budgeting Portfolios","authors":["Martin Arnaiz Iglesias","Adil Rengim Cetingoz","Noufel Frikha"],"paper_date":"2024-11-19","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2411.12323v1","page_url":"https://thequant.space/flowcharts/mirror-descent-algorithms-for-risk-budgeting-portfolios/","scores":{"empirical_rigor":6,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes deterministic and stochastic mirror descent (MD) algorithms, including a tamed gradient version, for computing risk budgeting portfolios across general risk measures.","quote":"We then introduce our deterministic MD (DMD) algorithm, providing con-vergence results along with a convergence rate for the weighted averaged sequence. Next, we propose and analyse a stochastic MD (SMD) scheme in a stochastic frame-work where the risk measure is expressed as the minimum of a convex map.","summary":"This paper introduces and examines numerical approximation schemes for computing risk budgeting portfolios associated to positive homogeneous and sub-additive risk measures. We employ Mirror Descent algorithms to determine the optimal risk budgeting weights in both deterministic and stochastic…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The learning step sequence satisfies \\(\\sum_{n}\\gamma_n = \\infty\\) and \\(\\sum_{n}\\gamma_n^2 < \\infty\\).","type":"statistical","stated":"explicit","quote":"Assume additionally that the learning step sequence (γn)n≥1 satisfies P n≥1 γn = ∞and P n≥1 γ2 n < ∞. (3.11) 11 Assume additionally that the learning step sequence (γn)n≥1 satisfies P n≥1 γn = ∞and P n≥1 γ2 n < ∞.","support":"full"},{"assumption":"The bound \\(m \\ge \\|y^*\\|_1\\) holds for the radius of the feasible set.","type":"model","stated":"explicit","quote":"Assume that m ≥∥y⋆∥1. Assume that m ≥∥y⋆∥1, that the regularity and integrability condi-tions on the function L of Proposition 3.1 are satisfied and that sup (ξ,y)∈R×(Rd +∩Bm) \b E[(∂ξL(ξ, −⟨y, X⟩))2] + E[X2 i (∂xL(ξ, −⟨y, X⟩))2] < ∞, 1 ≤i ≤d.","support":"full"},{"assumption":"Risk measures are positive homogeneous and sub‑additive.","type":"model","stated":"explicit","quote":"This paper introduces and examines numerical approximation schemes for computing risk budgeting portfolios associated to positive homogeneous and sub-additive risk measures.","support":"full"},{"assumption":"The stochastic risk measure can be expressed as the minimum of a convex map.","type":"model","stated":"explicit","quote":"Next, we propose and analyse a stochastic MD (SMD) scheme in a stochastic frame-work where the risk measure is expressed as the minimum of a convex map.","support":"full"},{"assumption":"The gradient of the value function may be unbounded, requiring a tamed version.","type":"other","stated":"explicit","quote":"By developing a tailored tamed version of the gradient, we intro-duced both deterministic and stochastic MD schemes, establishing their convergence along with a convergence rate for the weighted averaged sequence.","support":"full"}]},"datasets":{"status":"available","items":[{"name":"S&P 500 daily returns","provider":null,"kind":"market-prices","access":"unknown","period":"Aug 2008 to Apr 2022","universe":"S&P 500 components","quote":"We randomly select d stocks from the S&P 500 components (as of Q2 2022) and obtain their daily returns from August 2008 to April 2022.","support":"full"},{"name":"Student‑t mixture simulated data","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"We then simulate 106 data points from the fitted model and apply the SMD, c-SGD, and t-SGD algorithms to the simulated dataset, analyzing convergence by comparing the results to the reference portfolio.","support":"full"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"none","findings":[{"finding":"Both deterministic and stochastic MD algorithms remain stable as dimensionality increases and outperform the standard SGD algorithm recently proposed in the literature.","kind":"performance","quote":"Our numerical results illustrate these theoretical findings, demonstrating that both algorithms remain stable as dimensionality increases and perform more effec-tively than the standard SGD algorithm recently proposed in the literature.","support":"full"},{"finding":"The stochastic MD (SMD) algorithm converges to the minimum faster than t‑SGD in the experiments.","kind":"performance","quote":"The results indicate that the SMD algorithm converges to the minimum faster than t-SGD.","support":"full"},{"finding":"Tamed methods (t‑SGD and SMD) consistently avoid divergence across all tested cases, unlike c‑SGD which is sensitive to learning‑rate choice.","kind":"performance","quote":"The figure suggests that, although c-SGD can occasionally achieve accurate results, its convergence is highly sensitive to the initial choice of the learning rate γ.5 In contrast, the tamed meth-ods exhibit notable robustness to hyperparameter selection, consistently avoiding divergence across all tested cases.","support":"full"},{"finding":"Theoretical convergence of the weighted averaged sequence is established with explicit non‑asymptotic rates for both deterministic and stochastic MD schemes.","kind":"theoretical","quote":"We employ Mirror Descent algorithms to determine the optimal risk budgeting weights in both deterministic and stochastic settings, establishing convergence along with an explicit non-asymptotic quantitative rate for the averaged algorithm. We establish the a.s. convergence of the SMD algorithm, along with a non-asymptotic a.s. convergence rate for the weighted average sequence.","support":"full"}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":19,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":2,"extracted_at":"2026-11-14T19:54:05Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}