Hey /sci/, /g/-tard here. Can any /neuro/ guy (researcher, doctor, enthusiast, w/e) help me answer a few questions I have regarding biological neural networks that arose when studying artificial neural networks (ANNs)?
1. In ANNs, connections between neurons have a variable weight. Do signals sent by neurons always have the same potential, or does it vary?
2. In ANNs, neurons can have a threshold that dictates whether they propagate a received signal or not, depending on the signal strength. As I understand, bio-neurons also have a threshold potential. Is that threshold variable, or is it constant for every neuron?
3. In ANNs, the neural structure (i.e. number of neurons and how they are connected) is usually constant. It is set before training and usually does not change after that (weights are what change). I realize that with humans this is not always the case. For example, infants must certainly have less neurons and therefore less connections than adults, but for a fully grown human:
3.1. Does the total number of neurons shift up and down, or just down?
3.2. Can new physical connections between neurons happen? And can old connections be broken?
4. The answers to the previous questions may already hint at this, but how does learning happen in the brain? What changes at a neural level?
5. In the brain, is there a distinction between learned skills and data/knowledge? For example: is my skill at driving a car "encoded" in the same way as my knowledge of ANNs or my childhood memories?
Feel free to answer how many questions you'd like, or just hang around to marvel at the genius that is nature.
1. In ANNs, connections between neurons have a variable weight. Do signals sent by neurons always have the same potential, or does it vary?
2. In ANNs, neurons can have a threshold that dictates whether they propagate a received signal or not, depending on the signal strength. As I understand, bio-neurons also have a threshold potential. Is that threshold variable, or is it constant for every neuron?
3. In ANNs, the neural structure (i.e. number of neurons and how they are connected) is usually constant. It is set before training and usually does not change after that (weights are what change). I realize that with humans this is not always the case. For example, infants must certainly have less neurons and therefore less connections than adults, but for a fully grown human:
3.1. Does the total number of neurons shift up and down, or just down?
3.2. Can new physical connections between neurons happen? And can old connections be broken?
4. The answers to the previous questions may already hint at this, but how does learning happen in the brain? What changes at a neural level?
5. In the brain, is there a distinction between learned skills and data/knowledge? For example: is my skill at driving a car "encoded" in the same way as my knowledge of ANNs or my childhood memories?
Feel free to answer how many questions you'd like, or just hang around to marvel at the genius that is nature.
