This preview shows page 1. Sign up to view the full content.
Unformatted text preview: i1 V This electric circuit is described by a relation i1 = V /R1 and i2 = V /R2 . In order to learn linear system with single input and two output, two linear neurons are necessary: V i1 i2 (d) An ANN to predict the voltage on a diode based on its input electric current.
i V A diod is a non linear circuit. As in previous examples, we need a single input neuron and a single linear output neuron. However, in order to deal with non-linear systems, we need to introduce at least one hidden layer of tan hyperbolic units (default choice). i V 2 In the four preceding example, the network can be trained by recording a table of input and output voltage and intensity. Case a to c can be trained using the gradient descent (linear network). Case d requires to use the back-propagation algorithm. (e) A classifier identifying adults based on their age. This is the simplest classification task where the input point can be separated into adult and children immediately by one value (around 18 years old). Because the separa...
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