{"paper_id":"2308.02049","title":"Portfolio Optimization in a Market with Hidden Gaussian Drift and Randomly Arriving Expert Opinions: Modeling and Theoretical Results","authors":["Abdelali Gabih","Ralf Wunderlich"],"paper_date":"2023-08-03","paper_type":"theoretical","paper_url":"https://arxiv.org/pdf/2308.02049v2","page_url":"https://thequant.space/flowcharts/portfolio-optimization-in-a-market-with-hidden-gaussian-drif/","scores":{"empirical_rigor":3,"math_complexity":9,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper proposes a portfolio optimization framework for power‑utility investors in a market with a hidden Gaussian mean‑reverting drift, where information arrives from returns and randomly timed noisy expert opinions; it derives Kalman‑filter estimates, formulates a risk‑sensitive stochastic contr","quote":"This paper investigates the optimal selection of portfolios for power utility maximizing investors in a financial market where stock returns depend on a hidden Gaussian mean reverting drift process. Information on the drift is obtained from returns and expert opinions in the form of noisy signals about the current state of the drift arriving randomly over time.","summary":"This paper investigates the optimal selection of portfolios for power utility maximizing investors in a financial market where stock returns depend on a hidden Gaussian mean reverting drift process. Information on the drift is obtained from returns and expert opinions in the form of noisy signals…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The drift process µ follows a Gaussian Ornstein‑Uhlenbeck dynamics with mean‑reversion speed κ, volatility σµ, and all eigenvalues of κ have positive real parts (stable matrix).","type":"model","stated":"explicit","quote":"Here κ ∈Rd×d, σµ ∈Rd×d2 and µ ∈Rd are constants such that all eigenvalues of κ have a positive real part (that is, −κ is a stable matrix), Σµ := σµσ⊤ µ is positive deﬁnite, and W µ is a d2-dimensional Brownian motion such that d2 ≥d. The parameters µ,κ,σµ are called mean-reversion level, the mean-reversion speed and volatility of µ, respectively.","support":"full"},{"assumption":"The initial drift µ0 is normally distributed with mean m0 and covariance q0, independent of the Brownian motions driving returns and the drift.","type":"model","stated":"explicit","quote":"The initial value µ0 is assumed to be a normally distributed random variable independent of W µ and W R with mean m0 ∈Rd and covariance matrix q0 ∈Rd×d assumed to be symmetric and positive semi-deﬁnite.","support":"full"},{"assumption":"The Brownian motions driving the return process and the drift process are independent.","type":"model","stated":"explicit","quote":"For the sake of simpliﬁcation and shorter notation we assume that the Wiener processes W R and W µ driving the return and drift process, respectively, are independent.","support":"full"},{"assumption":"Expert opinions arrive at random times modeled as the jump times of a homogeneous Poisson process.","type":"data","stated":"explicit","quote":"The arrival dates are modeled as the jump times of a homogeneous Poisson process.","support":"full"},{"assumption":"Investors observe only the return process and the noisy expert signals; they do not observe the drift µ or the Brownian motion driving returns.","type":"market","stated":"explicit","quote":"2.2 Expert Opinions We assume that investors do not have a complete access to the market information, they can instead observe the historical data of the return process R but they neither observe the factor process µ nor the Brownian motion W R. Further, investors are aware of the model parameters such as σR,κ,µ,σµ and the distribution N (m0,q0) of the initial value µ0. Accordingly, information about the drift µ can","support":"full"},{"assumption":"Between expert‑opinion arrival dates the conditional covariance of the filter is deterministic, making the filter mean a diffusion process.","type":"model","stated":"explicit","quote":"In that setting the above conditional mean M is a diffusion process and the conditional covariance Q is deterministic.","support":"full"}]},"datasets":{"status":"none_found","items":[],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"none"},"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":"A dynamic programming equation for the risk‑sensitive control problem is derived and regularized to become strictly elliptic, enabling classical verification arguments.","kind":"theoretical","quote":"Then the dynamic programming equation associated with the regularized optimiza-tion problem is strictly elliptical so that classical veriﬁcation results can be applied.","support":"full"},{"finding":"Solutions of the regularized state equations converge in mean‑square to the solution of the original non‑regularized state equation as the perturbation parameter tends to zero.","kind":"theoretical","quote":"For this family of perturbed control problems, we ﬁrst show that for scaling parameters approaching zero, the solutions of the regularized state equations converge in the mean-square sense to the solution of the original, non-regularized state equation.","support":"full"},{"finding":"Reward functions of the regularized control problems converge uniformly to those of the original problem for all admissible strategies.","kind":"theoretical","quote":"Next, we show that the reward functions of the control problems converge uniformly for all admissible strategies.","support":"full"},{"finding":"Value functions of the regularized problem converge to the original value function, implying that regularized optimal strategies are ε‑optimal for the original problem.","kind":"theoretical","quote":"Finally, we prove that value functions of the regularized problem converge to the value function of the original problem. That result implies that optimal strategies for the regularized problem are nearly or ε-optimal in the original problem which gives a method for computing ε-optimal strategies.","support":"full"},{"finding":"The theoretical results are illustrated by extensive numerical experiments solving the associated PIDE, confirming the practical computation of ε‑optimal strategies.","kind":"theoretical","quote":"We ﬁnally point out that the theoretical ﬁndings of this paper are illustrated in [24] by results of extensive numerical experiments based on the numerical solution of the PIDE appearing in the dynamic programming equation. Outline.","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":16,"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":"2026-11-14T20:01:46Z"},"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/"}