NTAFTAWT User Guide
Last update: 12-27-2007This program feeds a set of patterns to the input of the neural network. The input pattern is a simple representation of seeing.
Example Use
- Start the "charpatt" program
- Use File/Open to open the file blockL1P2.txt
- Move the mouse over the neurons to see how their dendrites match the input pattern.
- Keep the iterations to 1, and press "Run", then move the mouse over neurons to see matches for the new pattern. Watching the same neuron over a few iterations should show that the number of dendrites that match the input pattern is increasing.
Interface
| Interface | Description |
| Left Scene | The left scene shows a representation of the eye rods/cones |
| Right Scenes | The right scenes shows a representation of the neurons. Moving the cursor over a neuron will highlight where the dendrites of the neuron are connected to the eye. The number of scenes on the right depends on the number of layers loaded by the configuration. Clikcing on a neuron will show all dendrites of the neuron of all layers. |
| Iterations | Each time the Run button is pressed, the fire and learn functions are performed for this many times. |
| Box/Character | Defines the type of input patterns that will be set to fill the network |
| History | This shows which neurons are firing over time. The gray scale indicates how many dendrites matched a pattern at some time. The colored dots indicate firing and remapping. A different history dialog will be loaded depending on which layer/scene the mouse was in last. |
| Quality | This shows which neurons fired at particular times and the neurons are sorted in order. This gives an indication of whether there are sets of neurons that uniquely identify a pattern. A different quality dialog will be loaded depending on which layer/scene the mouse was in last. |
| Run | This will run a certain number of iterations and perform
the following:
|