Implicit regularization in AI meets generalized hardness of approximation in optimization - Sharp results for diagonal linear networks.
Johan S. WindVegard AntunAnders C. HansenPublished in: CoRR (2023)
Keyphrases
- optimization algorithm
- sparse approximation
- global optimization
- conjugate gradient algorithm
- closed form
- artificial intelligence
- approximation algorithms
- quadratic programming
- error tolerance
- high quality
- learning theory
- risk minimization
- linear systems
- knowledge representation
- expert systems
- highly non linear
- social networks
- semidefinite
- ai systems
- machine learning
- polynomial kernels
- constrained optimization
- optimization method
- information theoretic
- computational complexity
- linear programming
- least squares