---
title: AP-PEN: A Hard-Constrained, Physics-Enforced Network for Convenience-Yield Inversion from Commodity Futures
url: https://www.ml-quant.com/papers/ssrn/7677598/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-11-06
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
identifier: SSRN 7677598
source_url: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7677598
featured: 2026-11-06
citations: unknown
topic: Derivatives & Volatility
---


# AP-PEN: A Hard-Constrained, Physics-Enforced Network for Convenience-Yield Inversion from Commodity Futures

A neural network that analytically enforces the Gibson-Schwartz PDE recovers commodity convenience yield from futures and avoided implausible estimates during the April 2020 WTI dislocation.

- Source: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7677598
- Identifier: SSRN 7677598
- Released: 2026-11-03
- First featured: Quant Letter No. 138 (2026-11-06): https://www.ml-quant.com/issues/2026-11-06/
- Citations (Semantic Scholar): not tracked
- Published in: not yet
- Topic: Derivatives & Volatility
- Authors: Evan Lynch

## Related

- [Deep Learning and GARCH Models for Financial Volatility](https://www.ml-quant.com/papers/ssrn/4589950/): A hybrid approach combining GARCH time series models with deep learning neural networks is proposed for forecasting financial volatility and risk, tested on S&P 500, gold, and Bitcoin prices.
- [Bermudan Commodity Options Pricing with Neural Networks](https://www.ml-quant.com/papers/repec/gam-jjrfmx-v-16-y-2023-i-3-p-192-d-1094945/): Multi-layered neural networks used for option pricing in commodity markets with high accuracy.
- [Neural Jumps for Option Pricing](https://www.ml-quant.com/papers/arxiv/2506.05137/): A neural jump stochastic differential equation model is proposed for option pricing, combining neural networks with the traditional jump diffusion model, enhancing pricing accuracy.
- [Applying Informer for Option Pricing: A Transformer-Based Approach](https://www.ml-quant.com/papers/doi/10-5220-0013320900003890/): The research uses the Informer neural network for option pricing in financial markets, showing its improved performance over traditional models and improving financial forecasting.
- [Model-Free Deep Hedging with Transaction Costs and Light Data Requirements](https://www.ml-quant.com/papers/arxiv/2505.22836/): The research shows that a neural network trained with just 256 trajectories can outperform the Black & Scholes formula and the Leland model in the Geometric Brownian Motion framework, indicating potential for real-time financial series application.
- [Unified GARCH-Recurrent Neural Network in Financial Volatility Forecasting](https://www.ml-quant.com/papers/arxiv/2504.09380/): The paper proposes a new GARCH-GRU model for financial volatility forecasting, showing better computational efficiency and forecasting accuracy than other models.