2015. június 14., vasárnap
Neural Networks mimic human brain, but is there something like backpropagation?
Neural Networks mimic human brain, but is there something like backpropagation? The method of backpropagation looks "too artificial", not something like brain can do. But in this case, what method is used by brain, and why don't we mimic it?
Feliratkozás:
Megjegyzések küldése (Atom)
I think we do not have enough knowledge about how human brain learns exactly. There are many parameters relevant to forming a synapse.
VálaszTörlésAirplane wings mimic bird wings.
VálaszTörlésJust a Joke ... "Neural Networks" do not mimics Nervous System Dynamics ... "Neural Networks" is a couple of words used about models, mechanical/statistical processes and procedures to simulate the mimics of some Peculiar Cases of Back-propagation Computing Models into Turing/Von Neuman devices...
VálaszTörlésAnd there are "more brain like" models? As I know neural networks are closest to the human brain.
VálaszTörlésOur brain cells learn through intensifying every connection that gets stimulated (This is what I know, it's probably more complicated than that).
VálaszTörlésThe neural networks in computer science are just inspired by the brain, they are not meant to be a 1:1 copy.
The "Delusion" by the Descartes way of thinking on Humans related to the Integration and dynamics of the environment into/out to the organism-environment made of the Brain the Object of Cult for the reductionist rituals of " Natural languages" and mathematical representations on some Apes ...
VálaszTörlésLaszlo Fazekas see http://research.ibm.com/cognitive-computing/neurosynaptic-chips.shtml
VálaszTörléshttp://www.research.ibm.com/articles/brain-chip.shtml
charles griffiths Yes, "Brain-like" is another stuff because is referred as the whole organ and the integrative recursion of information on a layer of modules, Hardware-software Integrative Dynamic Architectures of incomes/processes/outcomes ...
VálaszTörlésBack-propagation is the fastest way to achieve an optimal mapping in a small network but you're correct that it is not biologically plausible. In larger, more complicated recurrent networks, the Hebb rule can be used to learn similar functions, and it is very biologically plausible.
VálaszTörlés