-regularized Neural Networks are Improperly Learnable in Polynomial Time.
Yuchen ZhangJason D. LeeMichael I. JordanPublished in: CoRR (2015)
Keyphrases
- neural network
- dnf formulas
- exact learning
- learning algorithm
- pattern recognition
- computational complexity
- artificial neural networks
- equivalence queries
- regularized least squares
- fuzzy logic
- least squares
- neural nets
- multilayer perceptron
- hypothesis spaces
- positive data
- feed forward
- worst case
- special case
- dnf formulae
- membership and equivalence queries
- boolean functions
- equivalence and membership queries
- statistical queries
- pac learning
- membership queries
- activation function
- back propagation
- objective function
- genetic algorithm