{"paper_id":"2404.11257","title":"Deep Joint Learning valuation of Bermudan Swaptions","authors":["Francisco Gómez Casanova","Álvaro Leitao","Fernando de Lope Contreras","Carlos Vázquez"],"paper_date":"2024-04-17","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2404.11257v1","page_url":"https://thequant.space/flowcharts/deep-joint-learning-valuation-of-bermudan-swaptions/","scores":{"empirical_rigor":8,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a deep learning framework that combines differential machine learning, Monte‑Carlo‑like training samples and joint (multi‑task) learning to price Bermudan swaptions.","quote":"More precisely, we smartly combine several sophisticated neural network‑based concepts like differential machine learning, Monte Carlo simulation-like training samples and joint learning to come up with an efficient numerical solution.","summary":"This paper addresses the problem of pricing involved financial derivatives by means of advanced of deep learning techniques. More precisely, we smartly combine several sophisticated neural network-based concepts like differential machine learning, Monte Carlo simulation-like training samples and…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The underlying IRS is spot start with 1‑year payment frequency and 10‑year expiry.","type":"market","stated":"explicit","quote":"The underlying IRS is spot start (starting today) with payment frequency 1 year (in both legs) associated to the tenor structure {T1, T2. . . . , TM}, and expiry in 10 years."},{"assumption":"Bermudan call dates coincide with IRS payment dates, first call 1 year after spot.","type":"market","stated":"explicit","quote":"For the Bermudan swaption (respectively for the Cancellable IRS), the first call date is 1 year after the spot date of the underlying IRS, and the frequency of callability is also 1 year, i.e., the call dates of the Bermudan swaption coincide with the payment dates of the IRS."},{"assumption":"When exercised, the payment on the exercise date is assumed already paid and excluded from future payments.","type":"model","stated":"explicit","quote":"When the Bermudan swaption is exercised, the payment corresponding to the exercise date is assumed to be already paid, i.e, it is not included in the series of future payments."},{"assumption":"The notional is set to one.","type":"model","stated":"explicit","quote":"The notional is assumed to be the unit."},{"assumption":"Training samples are generated by Monte Carlo simulations with each path using a different set of model parameters.","type":"data","stated":"explicit","quote":"we simulate each of the Monte Carlo paths with a different set of parameters, such that, every single sample represents a very noisy price for that particular setting."},{"assumption":"The short‑rate dynamics follow the Linear Gauss‑Markov (LGM) model.","type":"model","stated":"explicit","quote":"In order to describe the time evolution of the over‑night interest rate, we employ the well‑known Linear Gauss Markov (LGM) model."}]},"datasets":{"status":"available","items":[{"name":"Monte Carlo simulated training set","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"an entire training set (with thousands or millions of samples) can be generated at the cost of a classical Monte Carlo simulation-based pricing method."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"synthetic"},"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":"pricing-calibration","findings":[{"finding":"Joint learning substantially reduces prediction variance compared with the plain DANN estimator.","kind":"performance","quote":"we again observe that the DANN trained relying on the joint learning approach provides more accurate estimations, significantly reducing the variance."},{"finding":"Increasing the number of Monte Carlo paths per sample halves the error, matching the theoretical n^{-1/2} convergence rate.","kind":"performance","quote":"when the number of paths is multiplied by four the error is approximately halved (according to the theoretical convergence rate of Monte Carlo methods, n−1/2)."},{"finding":"Systematically increasing the number of training samples improves predictions, but the reduction in inter‑quantile interval is limited.","kind":"performance","quote":"systematically increase the number of samples provided to the DANN improves the predictions, although the reduction of the interquantile intervals, i.e., in the deviations’ variance, is rather limited."},{"finding":"For valuation at future dates, joint learning reduces confidence intervals by about 3 basis points, with an average accuracy gain of 1 basis point.","kind":"performance","quote":"the DANN including joint learning leads to improved estimations, achieving a reduction in the confidence intervals of around 3 basis points. In average, the gain in accuracy reaches 1 basis point."},{"finding":"When most differences fall below ±3 basis points, further increases in samples or Monte Carlo paths yield negligible improvement.","kind":"negative","quote":"When most of the differences fall below ±3 basis point, a certain level of saturation is observed meaning that considering either more samples or more Monte Carlo paths per sample no longer reduces the deviations in the predictions (or the reduction results to be negligible)."}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":34,"citations":{"cited_by_in_corpus":1,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":1},"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":1,"extracted_at":"2026-11-14T18:11:02Z"},"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/"}