I have programmed a few neural network programs, but if you think about it, the human bran has a maximum number of neurons (in that you don't create new neurons when you learn something, you create new pathways to the different neurons withing the brain). You can, however, destroy neurons within the brain, thereby losing memories. One thing you have to consider, since there is a finite number of neurons, the "memories" have to manipulate the data in the neurons according to the pathways formed between the neurons. Therefore, it seems that possibly, each neuron contains somewhat of a fuzzy set within itselt in order to retain information. If you think about it, in a neural network, a node can have a value of 0.3452432545 for several linked memories, but if a new memory is formed with that pathway, then the value may change slightly to something like 0.3452432547 for the new pathway (which is part of many other pathways for the memory). Granted, you cannot really give a value for the state a neuron is in at any given moment, but think of it along those concepts. Theorhetically, the brain could hold an infinite amount of data, but you have to consider when you manipulate a neuron's "state", then it will make minor changes to other memories thereby causing degredation in another memory. Also, it is highly likely that the pathways play a role in the memories (obviouslly?) in terms of what is stored. Therefore, since pathways are ALWAYS formed, then it could be said that the brain could hold infinite information. But logically, that is not the case. Anyway.. that's my 3 pesos. Yeah yeah, go ahead and flame me already.:P
If someone does determine that this is possible, PLEASE contact me. The first thing I thought about was 4 400s. I have the same Tyan board w/2 Celeron 400s, and if I could get 2 more, I would be extremely happy (kernel compilation in 2 min)!!
Slashdot Mirror
I have programmed a few neural network programs, but if you think about it, the human bran has a maximum number of neurons (in that you don't create new neurons when you learn something, you create new pathways to the different neurons withing the brain). You can, however, destroy neurons within the brain, thereby losing memories. One thing you have to consider, since there is a finite number of neurons, the "memories" have to manipulate the data in the neurons according to the pathways formed between the neurons. Therefore, it seems that possibly, each neuron contains somewhat of a fuzzy set within itselt in order to retain information. If you think about it, in a neural network, a node can have a value of 0.3452432545 for several linked memories, but if a new memory is formed with that pathway, then the value may change slightly to something like 0.3452432547 for the new pathway (which is part of many other pathways for the memory). Granted, you cannot really give a value for the state a neuron is in at any given moment, but think of it along those concepts. Theorhetically, the brain could hold an infinite amount of data, but you have to consider when you manipulate a neuron's "state", then it will make minor changes to other memories thereby causing degredation in another memory. Also, it is highly likely that the pathways play a role in the memories (obviouslly?) in terms of what is stored. Therefore, since pathways are ALWAYS formed, then it could be said that the brain could hold infinite information. But logically, that is not the case. Anyway.. that's my 3 pesos. Yeah yeah, go ahead and flame me already. :P
If someone does determine that this is possible, PLEASE contact me. The first thing I thought about was 4 400s. I have the same Tyan board w/2 Celeron 400s, and if I could get 2 more, I would be extremely happy (kernel compilation in 2 min)!!