Thursday, April 17, 2008

Threshold

A big question is what should the threshold be? If it's 1, then this simple circuit (pictured left) will continually pulse. If the threshold is higher, then we only need to add more feed-back neurons to make it pulse. This makes me question wither or not there is any need for a higher threshold. One reason to have a higher threshold is so that a single neuron can't fire another neuron, unless it fires in very rapid succession. This would have a filtering effect only causing the second neuron to fire when there was a lot of activity going on. I think this is a good reason to have higher thresholds. But what should the threshold be? Because this is an evolutionary project, I want to think that this should be an evolved value. This way I hope the network will figure out the best value on it's own. This would also open the possibility for neurons that have different thresholds within the same system. I don't know if this is a good thing or a bad thing, but I figure the evolution will eventually take care of that. So to add to my specs of the Neurons
  • Threshold is evolutionary
  • Voltage is evolutionary
  • Threshold rate is evolutionary
  • If the threshold is met within the threshold rate, the neuron fires the voltage to all of it's listeners.

Define

I was going to translate my notes and put them up on the blog but, I'm lazy. So I'm just going to put the relevant stuff up so I can continue on. I'v been thinking about how I will create this neuron program. I've run into a problem when I neuron has to receive multiple signals at the same time. After thinking about how I could achieve this, I decided that it would be better to forgo the balance neuron idea and create neurons that mimic their real life counterparts. So here is my new model
  • Neurons have a threshold, if that threshold is met within a certain amount of time, the neuron fires.
  • Neurons always fire the same signal to all of it's listeners. (example: Neuron A has two receivers, both will recive a value of 1 when the neuron fires.)
  • Neurons can listen to as many other neurons as they wish.
  • Neurons can only listen to the same neuron once. (example: Neuron A listens to Neuron B, Neuron A can't not add another listener to Neuron B)

Sunday, April 6, 2008

Threshold

Real neurons have a threshold that must be met before they will fire. Should the balance neurons have a threshold? Without any threshold it would propagate like the image below.

Inital Idea

The human body is always struggling to find a balance. Not just a balance within it's self (like the regulation of glucose in the blood stream) but also within our vision (Images that aren't balanced cause distress while images that are in balance make us feel peaceful and relaxed) and within our lives. So my thought was that when trying to create new life, it should also struggle for balance. The easiest way to create a life form that struggles for balance is to have the very elements that make up that life form struggle for balance themselves.

Balance Neuron

A neuron that is always trying to balance it's input with it's output. If it gets a 5 as in input, it will return -5 as it's output. Sum all inputs, and invert the value for the output. Example: 2 + 3 + 2 => -7

By it's self, this might not seam like much, but chained together I hope to create a system that can solve new problems without human interaction. Maybe even learn to communicate with humans in an intelligent way.