Leveraging augmented-Lagrangian techniques for differentiating over infeasible quadratic programs in machine learning.
Antoine BambadeFabian SchrammAdrien B. TaylorJustin CarpentierPublished in: ICLR (2024)
Keyphrases
- augmented lagrangian
- quadratic program
- convex optimization
- machine learning
- constrained optimization
- total variation
- constrained optimization problems
- image denoising
- linear constraints
- regularization term
- linear programming problems
- denoising
- primal dual
- learning algorithm
- lagrange multipliers
- transfer learning
- linear program
- linear programming
- search space
- objective function
- training data
- support vector machine
- feature space
- feature selection