|
Simple Example
: Math Functions (cosine data) - The cosine function is a very good example of the kind of pattern that Psychonet is best at picking up. The values produced by this function range from 0 to 1 and then back and make a pattern that is clearly identifiable on a graph.
Setup
- Collect sample data from whatever means you'd like (500 elements or so) into text file. (Note: This file is provided as part of the download in the Examples folder)
- Use the New (Wizard) menu option to create your network.
- In the Preferences panel under "Fitness Function" tab set Tolerance to .1
Training
- Hit the Training button and then select your Sample Data file.
- Begin Training, when Current Error value approaches zero, stop training.
- Switch Info Panel to Fitness Function and look at the value. Is the network trained to 100%? 80%? Is this satisfactory? If not train more.
- Exit Training and save network when ready.
Processing
- To create a Processing Data set, choose the Processing Data Manager menu option and select "From Raw Data" button. This will quickly generate a Processing Data set based on the last values in your raw data
file.
- Once the Processing Data set is made and in the buffer, you can process the data with the "Process" menu option
|
|
