First page Back Continue Last page Overview Graphics
Problems with MLP
Choosing the number of hidden nodes
- how many are enough?
- Usually, the number of connections in the network should be less than the number of training examples
As number of connections approaches the number of training examples, generalisation decreases