Training

Section 3) Training the network

In order to predict values, Psychonet must be trained. It does this by taking portions of your sample data (called sample sets) and processing it.

Sample set definition : Sample sets are defined to be a collection of data as well as the correct value(s) that the neural net should arrive at after processing. The number of elements in the sample set must match the number of inputs+ the number of outputs in your network.

Each time the data is processed, the network will evaluate itself and determine how far off it was. It will then attempt to correct itself and repeat the test. This process will continue until it reaches the error threshold you have set or until you tell it to stop.

Over many iterations the network will gradually learn how to make predictions in your data based on any patterns that it finds.

Training can involve extraordinary amounts of time depending on the size of you data file and the speed of your computer. Due to this Psychonet allows you to save a trained network that you can load on demand.