{"paper_id":"2501.09727","title":"Convergence of a Deep BSDE solver with jumps","authors":["Alessandro Gnoatto","Katharina Oberpriller","Athena Picarelli"],"paper_date":"2025-01-16","paper_type":"theoretical","paper_url":"https://arxiv.org/pdf/2501.09727v1","page_url":"https://thequant.space/flowcharts/convergence-of-a-deep-bsde-solver-with-jumps/","scores":{"empirical_rigor":3.5,"math_complexity":9.2,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper proposes a deep learning‑based algorithm for solving decoupled forward‑backward SDEs with jumps and derives a priori and a posteriori error estimates for both finite and infinite jump activity cases.","quote":"We study the error arising in the numerical approximation of FBSDEs and related PIDEs by means of a deep learning-based method. Our results focus on decoupled FBSDEs with jumps and extend the seminal work of HAn and Long (2020) analyzing the numerical error of the deep BSDE solver proposed in E et al. (2017).","summary":"We study the error arising in the numerical approximation of FBSDEs and related PIDEs by means of a deep learning-based method. Our results focus on decoupled FBSDEs with jumps and extend the seminal work of HAn and Long (2020) analyzing the numerical error of the deep BSDE solver proposed in E et…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The underlying probability space is filtered, supports a standard Rd‑valued Brownian motion and a Poisson random measure with Lévy measure ν satisfying ν({0})=0 and integrability conditions on |z|².","type":"model","stated":"explicit","quote":"Setting and Preliminaries Given a ﬁxed time horizon T < ∞, let (Ω, F, F, P) be a ﬁltered probability space with the ﬁltration F = (Ft)t∈[0,T] satisfying the usual conditions. We assume that the ﬁltered probability space supports an Rd-valued standard Brownian motion W = (Wt)t∈[0,T] and a Poisson random measure N with associated L´evy measure ν, such that ν({0}) = 0, Z Rd(1 ∧|z|2)ν(dz) < ∞, and Z |z|≥1 |z|2ν(dz) < ∞.","support":"full"},{"assumption":"The coefficients b, σ, Γ, f, g satisfy the regularity and growth conditions collected in Assumption 2.1 (A1) and (A3).","type":"model","stated":"explicit","quote":"Note that if the function Γ : Rd × Rd →Rd satisﬁes Assumption 2.1 (A3) the following estimates hold for x ∈Rd |γ(x)| ≤|γ(0)| + Kγ|x| ≤eCγ(1 + |x|), |γ(x)|2 ≤2|γ(0)|2 + 2K2 γ|x|2 ≤Cγ(1 + |x|2) with eCγ := 2(|γ(0)| ∨Kγ) and Cγ := 2(|γ(0)|2 ∨K2 γ). Let Assumptions 2.1 (A1), (A3), 3.1 and 3.3 hold.","support":"full"},{"assumption":"Neural network approximations Uρn satisfy measurability, growth and Lipschitz‑type gradient bounds (Assumption 3.3).","type":"other","stated":"explicit","quote":"We assume that for each n = 0, ..., M −1 the function Uρn n : Rd →Rd is measurable and satisﬁes the following growth condition (3.6) Uρn n (x) 2 ≤An + Bn|x|2 for constants An, Bn≥0. Moreover, for each n = 0, ..., M −1 the function Uρn n : Rd →Rd is diﬀerentiable and the gradient DxUρn n (x) satisﬁes the following growth condition (3.7) DxUρn n (x) 2 ≤Cn for a constant Cn≥0. Remark 3.4.","support":"partial"}]},"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 priori error estimates are obtained for the deep BSDE solver in both finite and infinite jump activity regimes.","kind":"theoretical","quote":"We provide a priori and a posteriori error estimates for the finite and infinite activity case. Section 4 provides a priori estimates both for the ﬁnite and inﬁnite activity case. 2.","support":"full"},{"finding":"A posteriori error estimates are derived, extending previous results and covering infinite activity jumps.","kind":"theoretical","quote":"Our analysis of a posteriori estimates (which excludes weak coupling) is in line with that of [30] but we also consider inﬁnite activity jumps and, more importantly, we obtain a priori estimates which are missing in [30]. In Section 3 we provide a self contained derivation of a posteriori estimates that is in line with the results of [30] but is extended to consider the inﬁnite activity case in Section 3.2.","support":"full"},{"finding":"Path‑regularity of the discretized solution is established, showing an overall L²‑error of order Δt.","kind":"theoretical","quote":"Then the following path regularity holds max n=0,...,M−1E \" sup t∈[tn,tn+1] Xt −ee Xt 2 # + max n=0,...,M−1 E \" sup t∈[tn,tn+1] Yt −eeY t 2 # + E \u0014Z T 0 Zt −eeZt 2dt \u0015 + E \u0014Z T 0 Ψt −eeΨt 2dt \u0015 ≤C∆t, (4.1) where C is a constant only depending on the data of the FBSDE in (2.3)-(2.4) and the processes ee X, eeY , eeZ, eeΨ are deﬁned on [tn, tn+1) by ee Xt := Xn, eeY t := Yn, eeZt := 1 ∆tE \u0014Z tn+1 tn Zsds Fn \u0015 ,","support":"full"}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"available","supports":[],"contradicts":[],"qualifies":[],"compared_with":12,"citations":{"cited_by_in_corpus":3,"by_independent_authors":2,"how_cited":{"used_as_baseline":0,"uses_method_or_data":1,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":2},"independent_papers":[{"paper":{"paper_id":"2605.06281","title":"INEUS: Iterative Neural Solver for High-Dimensional PIDEs","paper_date":"2026-05-07","authors":["Jean-Loup Dupret","Davide Gallon","Patrick Cheridito"],"empirical_rigor":6,"page_url":"https://thequant.space/flowcharts/ineus-iterative-neural-solver-for-high-dimensional-pides/"},"how":"uses_method_or_data","quote":null}],"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":"2026-11-14T19:52:56Z"},"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/"}