Lookback Option Pricing under VG and NIG Models: Label Quality and Shape-Constrained Neural Surrogates

This repository contains the derived calibration outputs, numerical results, figures, and supporting reproducibility materials associated with the study “Lookback Option Pricing under VG and NIG Models: Label Quality and Shape-Constrained Neural Surrogates.” The archived results include regime-balanced SPX calibration summaries for twelve market dates, calibration-robustness analyses, derived VG and NIG parameter estimates, lookback-pricing results, model-by-regime summaries, surrogate-evaluation outputs, and numerical diagnostics used in the study. The empirical analysis is based on 47,822 raw option quotes, from which 6,000 contracts were retained for calibration after applying the documented filtering and reconstruction procedures. The repository also includes numerical results for the calibrated lookback-pricing experiments, including 108 evaluated cases per model, together with figures and supplementary numerical-audit material. Raw historical SPX option-chain data are not redistributed because they are subject to the data provider’s licensing terms. The repository instead provides the derived results and documentation needed to identify the market dates, filtering rules, calibration procedures, and reported numerical outputs. The corresponding source code and trained neural-network checkpoints are archived separately on Zenodo at DOI: 10.5281/zenodo.22953623.