🎯 Awesome Decision-Focused Learning

A curated list of Decision-Focused Learning (DFL) papers, code, and resources.

Bridging the gap between prediction and optimization.

Overview • Papers • Libraries • Contributing

---

📖 Overview

Decision-Focused Learning (DFL) is an emerging paradigm that integrates machine learning with downstream optimization tasks. Unlike traditional two-stage approaches that minimize prediction error, DFL directly minimizes decision regret — the suboptimality of decisions made using predicted parameters.

---

📚 Papers

📐 Differentiable Optimization Layers

Embedding optimization problems as neural network layers with exact or approximate gradients.

Year	Venue	Paper	Keywords	Code
2017	ICML	OptNet: Differentiable Optimization as a Layer in Neural Networks	QP KKT
2020	NeurIPS	Interior Point Solving for LP-based prediction+optimisation	LP Interior Point
2024	NeurIPS	BPQP: A Differentiable Convex Optimization Framework for Efficient End-to-End Learning	QP ADMM Efficient
2025	NeurIPS	Differentiation Through Black-Box Quadratic Programming Solvers	QP Black-box Modular

📉 Surrogate Loss Methods

Designing tractable loss functions that provide informative gradients for training.

Year	Venue	Paper	Keywords	Code
2017	NeurIPS	Task-based End-to-end Model Learning in Stochastic Optimization	Stochastic End-to-End
2019	AAAI	Melding the Data-Decisions Pipeline: Decision-Focused Learning for Combinatorial Optimization	CO QP Smoothing
2021	Management Science	Smart "Predict, then Optimize"	SPO+ Convex Surrogate
2021	IJCAI	Contrastive Losses and Solution Caching for Predict-and-Optimize	NCE Caching Efficient	-
2022	NeurIPS	Decision-Focused Learning without Differentiable Optimization: Learning Locally Optimized Decision Losses	LODL Learned Loss
2024	NeurIPS	Decision-Focused Learning with Directional Gradients	PG Loss Zeroth-order	-
2025	NeurIPS	Solver-Free Decision-Focused Learning for Linear Optimization Problems	LAVA Solver-Free
2025	arXiv	Minimizing Surrogate Losses for Decision-Focused Learning using Differentiable Optimization	DYS-Net Surrogate	-

🎲 Perturbation & Gradient Approximation

Using randomization or geometric insights to approximate gradients through discrete solvers.

Year	Venue	Paper	Keywords	Code
2020	NeurIPS	Learning with Differentiable Perturbed Optimizers	PFYL Perturbation Fenchel-Young	-
2022	ICML	Decision-Focused Learning: Through the Lens of Learning to Rank	LTR Ranking
2023	ICLR	Backpropagation through Combinatorial Algorithms: Identity with Projection Works	IWP Projection Simple

📊 Surveys & Benchmarks

Year	Venue	Paper	Description
2024	EJOR	A Survey of Contextual Optimization Methods for Decision-Making under Uncertainty	Comprehensive survey covering DFL, contextual optimization, and stochastic programming

---

🛠️ Libraries & Benchmarks

Library	Description	Links
PyEPO	PyTorch-based End-to-End Predict-then-Optimize Library

---

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.

Ways to contribute:

• 📝 Add new papers
• 🔗 Update links and code repositories
• 📖 Improve descriptions
• 🐛 Report issues

---

⭐ Star this repo if you find it useful!

Made with ❤️ for the DFL community