Research map / Architectures and efficiency / RNNs and normalization
Leveraging augmented-Lagrangian techniques for differentiating over infeasible quadratic programs in machine learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a differentiable augmented-Lagrangian-based approach for differentiating through quadratic programs that may be infeasible, enabling optimization layers to train robustly.
Problem
Optimization layers in neural networks can become infeasible during training, and standard differentiation through quadratic programs breaks down in that case; this matters for learning, control and robotics applications.
Approach
Uses augmented-Lagrangian techniques to make the differentiation of quadratic programs well-defined even when the underlying problem is infeasible, so optimization layers can be trained through infeasible instances.
Experimental setup
Builds on: OptNet cvxpylayers
Closest papers by meaning
- Differentiation Through Black-Box Quadratic Programming Solvers (NeurIPS 2025 · similarity 0.89)
- Principled Data Augmentation for Learning to Solve Quadratic Programming Problems (NeurIPS 2025 · similarity 0.89)
- Deep Distributed Optimization for Large-Scale Quadratic Programming (ICLR 2025 · similarity 0.89)
- Expressive Power of Graph Neural Networks for (Mixed-Integer) Quadratic Programs (ICML 2025 · similarity 0.88)
- Predicting Lagrangian Multipliers for Mixed Integer Linear Programs (ICML 2024 · similarity 0.87)
- Learning Provably Improves the Convergence of Gradient Descent (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.