{"paper_id":"2412.12213","title":"The AI Black-Scholes: Finance-Informed Neural Network","authors":["Amine M. Aboussalah","Xuanze Li","Cheng Chi","Raj Patel"],"paper_date":"2024-12-15","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2412.12213v1","page_url":"https://thequant.space/flowcharts/the-ai-black-scholes-finance-informed-neural-network/","scores":{"empirical_rigor":3,"math_complexity":7.5,"quadrant":"Lab Rats"},"proposes":{"status":"available","statement":"The paper proposes a Finance-Informed Neural Network (FINN) for option pricing and hedging that integrates financial theory directly into machine learning.","quote":"We propose a Finance-Informed Neural Network (FINN) for option pricing and hedging that integrates financial theory directly into machine learning.","summary":"In the realm of option pricing, existing models are typically classified into principle-driven methods, such as solving partial differential equations (PDEs) that pricing function satisfies, and data-driven approaches, such as machine learning (ML) techniques that parameterize the pricing function…","strategy":"Finance-Informed Neural Network (FINN) delta hedging"},"assumptions":{"status":"available","items":[{"assumption":"The underlying asset follows a generic diffusion with drift μ and volatility σ satisfying Assumption 3.1.","type":"model","stated":"explicit","quote":"where 𝑊𝑡 is a standard Brownian motion under the risk-neutral measure Q, and 𝜇, 𝜎 satisfy Assumption 3.1."},{"assumption":"The hedge ratios are well‑defined under Assumption 3.2.","type":"model","stated":"explicit","quote":"which are well-defined under Assumption 3.2."},{"assumption":"Risk‑free rate r and dividend yield q are fixed from market proxies.","type":"market","stated":"explicit","quote":"with the risk‑free rate 𝑟 and dividend yield 𝑞 fixed from market proxies."},{"assumption":"The replication loss is minimized in continuous‑time limit, implying vanishing replication error leads to arbitrage‑free pricing.","type":"other","stated":"explicit","quote":"vanishing replication loss implies convergence of both prices and Greeks to the arbitrage‑free solution on compact domains."}]},"datasets":{"status":"available","items":[{"name":"Historical spot returns","provider":null,"kind":"market-prices","access":"unknown","period":"2010–2016","universe":null,"quote":"the model under the physical measure P using spot returns from 2010–2016 by matching empirical moments"},{"name":"Training option sample","provider":null,"kind":"options","access":"unknown","period":null,"universe":null,"quote":"on the training option sample via least‑squares fitting of implied volatilities"},{"name":"Synthetic GBM price paths","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"we simulate 50,000 independent GBM price paths using the testing year’s volatility"}],"known":[],"sample":{"start":null,"end":null,"market":"SPY options","asset_class":"options","frequency":"daily"},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"Costs δ ∈ {0, 0.01, 0.02} are applied to the traded amount when rebalancing the underlying stock.","quote":"cost 𝛿∈{0, 0.01, 0.02} is applied to the traded amount. The cash account therefore evolves according to …","source":"evidence_card","rule":"R2","candidates":{"front_matter":"unclear","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Costs δ ∈ {0, 0.01, 0.02} are applied to the traded amount when rebalancing the underlying stock.","card_quote":"cost 𝛿∈{0, 0.01, 0.02} is applied to the traded amount. The cash account therefore evolves according to …","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"pricing-calibration","findings":[{"finding":"FINN accurately recovers classical Black‑Scholes prices and remains stable in stochastic volatility (Heston) settings where analytical Greeks are unavailable.","kind":"performance","quote":"FINN accurately recovers classical Black–Scholes prices and performs robustly in stochastic volatility environments, including the Heston model"},{"finding":"FINN reduces out‑of‑sample implied‑volatility MAE from 0.1153 to 0.0456 (60.5% improvement) versus a hybrid Heston model.","kind":"pricing-calibration","quote":"FINN reduces the mean absolute error (MAE), from 0.1153 to 0.0456, corresponding to a 60.5% improvement"},{"finding":"Delta‑hedging with FINN yields lower variability and tail risk than BSM‑Delta, especially when transaction costs of 1% or 2% are present.","kind":"performance","quote":"FINN–Delta consistently exhibits lower variability and smaller downside risk"},{"finding":"Put‑call parity emerges endogenously without being imposed as an explicit constraint.","kind":"theoretical","quote":"adherence to put‑call parity provides a strong out‑of‑sample test of whether the model has internalized the economic structure"},{"finding":"The replication‑error loss (mean‑squared deviation from risk‑free growth) drives the learning of economically meaningful sensitivities.","kind":"theoretical","quote":"FINN trains 𝜃 by minimizing the mean squared deviation from risk‑free growth: L(𝜃) = E …"}],"reported_performance":{"strategy":"Finance-Informed Neural Network (FINN) delta hedging","sharpe":null,"sharpe_benchmark":null,"annual_return_pct":null,"total_return_pct":null,"max_drawdown_pct":null},"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":42,"citations":{"cited_by_in_corpus":1,"by_independent_authors":1,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":1,"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":1,"extracted_at":"2026-11-14T17:49:54Z"},"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/"}