Network Architectures Improve Option Pricing Accuracy
Researchers empirically compare neural network architectures for supervised learning of option prices and implied volatilities, evaluating Black‑Scholes and Heston models and real market data. They find generalized highway networks achieve the best mean-squared error and training-time trade-off for Black‑Scholes and Heston pricing, while a simplified DGM variant yields lowest error for transformed implied‑volatility tasks; capacity‑normalized comparisons are included.
Key Points
- 1Compare neural architectures on supervised option pricing and implied-volatility learning for Black‑Scholes and Heston models.
- 2Identify generalized highway networks deliver lowest MSE and fastest training for Black‑Scholes and Heston experiments.
- 3Recommend practitioners choose highway networks for price approximation and simplified DGM variants for implied‑volatility estimation.
Scoring Rationale
Practical empirical comparison offering clear architecture guidance, limited by a niche finance focus and being a single arXiv preprint.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
