Skip to content

Operator Deep Smoothing for Implied Volatility

Bibliographic record. Follow the original-source link for the publication.

Field Value
Primary domain Volatility
Other domains
Methods Financial Ml
Facets
Authors Ruben Wiedemann, Antoine Jacquier, Lukas Gonon
Published 2025-06-16
Source arXiv Quantitative Finance History
Identifiers arxiv:2406.11520
URL Open original source

Editorial synthesis

Why it matters

The paper proposes neural-operator based implied-volatility nowcasting, which could affect throughput and consistency of volatility-surface generation in high-frequency option data, but evidence is currently abstract-level assertion. (abstract:S1, abstract:S2, abstract:S3, abstract:S4, abstract:S5, abstract:S6, abstract:S7, abstract:S8, abstract:S9, abstract:S10)

Main author claims

  • The authors claim an operator deep smoothing approach that maps observed data directly to a smoothed implied-volatility surface, contrasting with the limitations of classical neural networks under dynamic spatial configurations. (abstract:S1, abstract:S2, abstract:S3, abstract:S4, abstract:S6)
  • They claim GNO-based training achieves high accuracy on ten years of raw intraday S&P 500 options with one model, while enforcing no-arbitrage constraints and showing robustness to input subsampling, including comparisons to NN and SVI. (abstract:S6, abstract:S7, abstract:S8, abstract:S9)

Data, method, or discussion scope

The scope includes framework claims, ten-year raw intraday S&P 500 coverage, and benchmark comparison mentions, but no explicit error metrics, compute budget, parameter scale, or bias decomposition is provided. (abstract:S1, abstract:S2, abstract:S6, abstract:S7, abstract:S8, abstract:S9, abstract:S10)

Main limitations

Terms like high accuracy and robustness are presented without error distributions, subsampling-bias controls, or explicit formulation of no-arbitrage constraints, limiting reproducibility of deployment consistency. (abstract:S7, abstract:S8, abstract:S9, abstract:S10)

Relationships

  • None recorded.