Vector-output ReLU Neural Network Problems are Copositive Programs: Convex Analysis of Two Layer Networks and Polynomial-time Algorithms.
Arda SahinerTolga ErgenJohn M. PaulyMert PilanciPublished in: ICLR (2021)
Keyphrases
- neural network
- optimization problems
- computational complexity
- quadratic optimization problems
- convex optimization problems
- learning algorithm
- computationally hard
- multi layer
- minimization problems
- computational problems
- benchmark problems
- variational inequalities
- desired output
- partial solutions
- approximation algorithms
- worst case
- social networks
- convex optimization
- convex functions
- convex relaxation
- convex hull
- np hardness
- convex programming
- combinatorial optimization
- newton method
- artificial neural networks
- linearly constrained
- genetic algorithm