Expert: George Moustris - 2/10/2004

Question
what is the difference between artificial neural networks and the neural network present in our mind?
does the machine cotaining these circuits resemble our mind in all respects?


Answer
Hello Lokesh

artificial neural networks (ANNs) try to capture some certain aspects of the processing a real neuron does. They're an abstract model of the real neuron, not an identical representation. ANNs are very very very much simpler than real neural networks, even so, complex enough. Chemical reactions are not modelled. Neither are neurotransmittors. Spatial processing in the core (soma) of the real neuron is ignored. Non-linear coupling effects are also ignored. The asynchronous way of biological neuron communication in modelled only in special ANNs (asynchronous ANNs). All other types are synchronous. The firing of the neurons is also very poorly modelled. Real neurons fire pulses (pulse trains), while ANNs have a continuous output. And again, only a special kind of ANNs model this behaviour (pulse-train ANNs). There are many other minor and major things that are ignored or modelled loosely. But even so, the final outcome ( the ANN) is very complex and difficult to study.
Few machines (i am aware of just one) contain neural networks. That is, a hardware implementation of ANN. ANNs are most oftenly simulated in software. And of course, the computer that either has a hardawre implementation of ANNs, or a software one, does not resemble the human mind in all aspects. They resemble it in some ways, but not all. To be more precise, they resemble some parts of the brain. But the analogy is very poor, since the brain contains about 10^30 neurons, and a software simulation models a dozen or tens of dozens neurons.

Best G.