This is the base value of the neuron, the zero point.
-- part contents for background part 29
----- text -----
This is where you enter the transformation function. It should be in the form: "put X into Y", where X be modified, like x+3 or (5*x)-3.
-- part contents for background part 30
----- text -----
This is the value the neuron gives to it's connections.
-- part contents for background part 31
----- text -----
This is where the neuron stores the values it recieves from other neurons.
-- part contents for background part 32
----- text -----
This is where the connections are listed. Connections are added using the "Connect" button. Connections are removed by deleting them from this field and removing thier corresponding strength.
-- part contents for background part 33
----- text -----
This is where the strengths for the connections are listed. They can be defined by simply typing the value in the corresponding row after the connection has been defined.
-- part contents for background part 34
----- text -----
The space is used to title a neuron.
-- part contents for background part 35
----- text -----
This displays the card id, which is useful in identifing connections.
When you create a new neuron, it is automatically entered.
-- part contents for background part 36
----- text -----
This button adds a new neuron to the stack.
-- part contents for background part 37
----- text -----
This clears ALL the Input and Output values from the network.
-- part contents for background part 38
----- text -----
This runs the network for a specified number of cycles.
-- part contents for background part 39
----- text -----
Press this button to enter the connection once you are at the neuron you want to connect to.
-- part contents for background part 40
----- text -----
Use this button to start the connection process. After pressing this button, move to the card you want to connect to and press the "Confirm Connect" buttton, or press "Cancel Connect" to cancel the connection.
-- part contents for background part 41
----- text -----
Press this button if you want to cancel the connecting process.
-- part contents for background part 42
----- text -----
This pops the last card in the stack.
-- part contents for background part 43
----- text -----
This finds a specified string.
-- part contents for background part 44
----- text -----
This goes to the next neuron.
-- part contents for background part 45
----- text -----
This goes to the prev. neuron.
-- part contents for background part 27
----- text -----
This stack is the documentation for the Neural Network Construction Stack (NNCS). The Neural Network Construction Stack allows a user to easily build and run a simulation of a neural network in Hypercard.
For "the rest of us", a neural network simulation is a collection of hypothetical neurons and the connections among them. The neurons pass information among themselves in much the same way that the neurons in our brain pass information. The neurons in our brain pass information as packets of chemicals called neurotransmitters. For simulation purposes we portray the amount and type of neurotransmitter that flows between neurons as numbers.
Each neuron in the NNCS can be characterized by a level of activity analogous to the firing rate of real neurons. The connection between two neurons (a synapse) is a one-way connection that can be characterized by a coupling strength. The coupling strength dictates how and how much of the activation of one neuron (the presynaptic neuron) is passed on to the other neuron (the postsynaptic neuron). In particulur, the amount of neurotransmitter passed across the synapse is determined as follows.
neurotransmitter =
presynaptic firing rate X
coupling strength.
The activity in the postsynaptic neuron is a function of the total amount of neurotransmitter that it receives and some preset resting level of activity. (This function is called the transformation function.) Synapses with positive coupling strengths are therefore excitatory since they add activation to the postsynaptic neuron. Synapses with negative coupling strengths are inhibitory since they subtract activation in the postsynaptic neuron.
The NNCS uses discrete cycles to represent time. In this system, each neuron has two activity levels. An input level that collects neurotransmitter on the current cycle, and an output cycle that represents firing rate on the previous cycle. The cycle of events on each cycle is as follows.
1. Presynaptic output levels are multiplied by coupling strengths and accumulated in postsynaptic input levels.
2. The transformation function is applied to the input value in each neuron.
3. For each neuron, the transformation result is placed in the output level.
Well, there's the basics of this simulation, use this help card to learn how to build your own neural network simulations. All you have to do is click on Any part of this card, and an explanation of it will come up. To show/hide the documentation click on the "Show Doc's" button. To make a simulator stack or to get an explanantion of the included sample, click on the "Info" button. Enjoy!
To print out these directions choose print report, with only the Neural Network Directions field selected, and the print card names disabled. To print out a Neural Network, choose Print Report with the print in rows option and the print card names disabled.
For questions, comments, or problems I may be reached at the following address:
Larry Halff
4918 N. 33rd Rd.
Arlington, VA 22207
GEnie: LHalff
CIS: 72250,2211
Although this stack is free, it is copyrighted and may not be sold or distributed in any commercial packeage without prior consent of the author (except for electronic bullten boards with a membership fee, or user's groups with a resonable duplication fee).
(1994)[Mac].iso/Science/Biology/NNCS/NNCS/card_2974_render.png){kind=link}
(1994)[Mac].iso/Science/Biology/NNCS/NNCS/card_4187_render.png){kind=link}