This preview shows page 1. Sign up to view the full content.
Unformatted text preview: tion can be done linearly, a single neuron can be used. In opposition to case a, the output function must the logistic sigmoid and the error function the cross entropy. Training can be done using a list of people with their age, and using the gradient descent algorithm. Age Adult (f) A cheese factory is starting a new cheese production. They have customers for strong mouldy cheeses and light clean cheeses, but strong clean cheeses and light mouldy ones are much less appreciated. They have a sensor deciding if a cheese is tasty (strong/light) and another one identifying the presence of mould. They want to use this input to identify cheeses that are ready to be sold. In this case there is no way to separate this data set using a single line in the (strength,mould) space. This is called the XOR problem. To solve this problem, a layer of hidden neuron must be introduced (with tanh activation). As in previous example, the error function must be the cross entropy and the output activation function must be the logistic sigmoid.
Mould Ready to sell
View Full Document
This note was uploaded on 03/21/2010 for the course IT ANN taught by Professor Mario during the Spring '10 term at École Normale Supérieure.
- Spring '10