{"paper_id":"2311.12239","title":"Quantum-inspired nonlinear Galerkin ansatz for high-dimensional HJB equations","authors":["Chuhao Sun","Asaf Cohen","James Stokes","Shravan Veerapaneni"],"paper_date":"2023-11-20","paper_type":"theoretical","paper_url":"https://arxiv.org/pdf/2311.12239v1","page_url":"https://thequant.space/flowcharts/quantum-inspired-nonlinear-galerkin-ansatz-for-high-dimensio/","scores":{"empirical_rigor":2,"math_complexity":9,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper introduces a family of nonlinearly parametrized trial functions with analytically tractable evolution equations as a precursor to Neural Galerkin methods for solving high‑dimensional time‑dependent HJB equations in finance.","quote":"This paper is inspired by the potential of solving Hamilton-Jacobi-Bellman (HJB) PDEs using Neural Galerkin methods and commences the exploration of nonlinearly parametrized trial functions for which the evolution equations are analytically tractable. As a precursor to the Neural Galerkin scheme, we present trial functions with evolution equations that admit closed-form solutions, focusing on time-dependent HJB","summary":"Neural networks are increasingly recognized as a powerful numerical solution technique for partial differential equations (PDEs) arising in diverse scientific computing domains, including quantum many-body physics. In the context of time-dependent PDEs, the dominant paradigm involves casting the…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The HJB equation admits a unique solution under standard regularity conditions on the data of the control problem.","type":"model","stated":"explicit","quote":"Under some conditions on the data of the problem, the value function V is the unique solution of the following PDE, called the HJB equation, [6] ( (∂tϕ)(t, x) + H(t, x, (Dxϕ)(t, x), (D2 xϕ)(t, x)) = 0, (t, x) ∈[0, T] × Rd, ϕ(T, x) = g(x), where Dxϕ is the gradient of ϕ and D2 xϕ is the Hessian matrix; the function H is the Hamiltonian associated with the problem, where in the case described above, is given by: H(t, x","support":"full"},{"assumption":"Asset dynamics follow a constant‑coefficient Black‑Scholes model with constant drift, volatility, and Sharpe ratio.","type":"market","stated":"explicit","quote":"Consider the dynamics where both S and Y follow constant coefficient Black-Scholes model: dSt = (r + λσ)Stdt + σStdBt, S0 = s0, dYt = b ⊙Ytdt + a ⊙YtdWt, Y0 = y0 > 0, where λ = µ −r σ . is the Sharpe ratio for the risky asset S, and µ, σ are constants.","support":"full"},{"assumption":"The trial functions are sufficiently regular (L2 integrable) so that the integrals defining M and V converge.","type":"model","stated":"explicit","quote":"In order for the integrals defining M and V to converge, by Cauchy-Schwarz inequality it suffices to require that eF(t, θ, ·) ∈L2(Ω) and ∂uθ ∂θi ∈L2(Ω) for all t ∈[0, T], θ ∈Rp+1 and 1 ≤i ≤p + 1. It is convenient (but not necessary) to also demand that uθ ∈L2(Ω) for all θ ∈Rp+1.","support":"full"}]},"datasets":{"status":"available","items":[{"name":"Synthetic 3‑dimensional grid experiment","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":"d = 3 (x, y dimensions)","quote":"The total space dimension including x, y is d = 3, and the space grid is [0, 4] along all dimensions. The default value of parameters are T = 1, γ = 0.5, r = 0.05, λ = 0.1, a = \u0014 0.3 0.0 0.0 0.3 \u0015 , b = [0.2, 0.2] and ρ = [0.1, 0.1].","support":"full"}],"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":"For the non‑tradable stock example, the required integrals can be performed in closed form, yielding explicit ODEs for the variational parameters.","kind":"theoretical","quote":"The necessary integrals can be performed in closed form and one finds the following results, M = α2 \u00141 0 0 1 4 \u0015 , V = α2 \u0014−1 2(λ2 + r) 1 4r \u0015 . Plugging the above expressions into (2.2), we find the following time dependence for the variational parameters, d dt \u0014log α log β \u0015 = \u0014 −1 2(λ2 + r) r \u0015 .","support":"full"},{"finding":"The NG method reproduces the exact solution of the PDE in this case, as verified by substituting the trial function back into the original equation.","kind":"theoretical","quote":"Remarkably, one finds that the NG reproduces the exact solution of the PDE in this case, which can be verified by plugging uθ(t) back into (2.1).","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":10,"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":"2026-11-14T20:00:13Z"},"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/"}