Multilayer feedforward networks with a nonpolynomial activation function can approximate any function.
Moshe LeshnoVladimir Ya. LinAllan PinkusShimon SchockenPublished in: Neural Networks (1993)
Keyphrases
- activation function
- multilayer feedforward
- backpropagation algorithm
- network size
- neural network
- artificial neural networks
- nonlinear functions
- back propagation
- hidden layer
- sigmoid function
- multilayer perceptron
- feed forward
- feedforward neural networks
- learning rate
- feed forward neural networks
- neural nets
- pattern recognition
- network structure
- recurrent neural networks
- basis functions
- small number
- training phase
- evolutionary algorithm
- support vector
- reinforcement learning
- decision making
- genetic algorithm