Lab 15: Processing with Neurons
- We're going to play with my neural net simulator, CANT.
- An old
tutorial might be worth a look.
- There's probably a better way to do this, but...
- In Eclipse Open up a new java project names CANT23.
Make sure that the execution JRE is JavaSE-1.6 (others may
work).
- Grab CANT3939.zip and stick it
somewhere.
- From the Eclpse File menu, import the archive file.
- Move all the src files up to CANT23/src/defaultpackage
- Move param.xml up to CANT23
- Now run the base class CANT23.
- You should get a window with a blue matrix and 400 cells.
- Each cell represents a neuron.
- If you widen the window, you can step or start.
- Try to get the neurons to persistently fire without external
activation, by setting parameters.
- Get the default 10x10 network running.
- Change the param.xml file so that you have 4 patterns with
5 neurons each.
- Add connections from those neurons to other neurons
- In CANT23 modify makeNewSystem to call a function
in CANTNet that sets up synapses.
- You might want to look at Iris3CANT to do this.
- Make a function stop in CANTNet with a print to
assure it's called.
- Now add connections (using the addConnection function),
from the input neurons (the ones from the pattern) to
other neurons.
- Make the weights large enough, or have enough connections
to make those to neurons fire.
- Make connections to make the toNeurons fire persistently.
Make the first pattern turn on one CA.
- Make the second pattern, in conjunction with the
persistent CA, turn on a second CA.
- Make the second CA, turn off the first. You can
have negative weights.
- Parse 3 sentences.