NeuralNetworks

Quora Answer : Is a neural network actually a good model of how the brain works?

Aug 21, 2018

By definition, a model tries to capture some useful properties of the thing it is modelling, but throws away other properties that it thinks aren't important.

So some neural networks are intended to be closer and more detailed models of how particular brains work. Others are high-level abstractions that only seek to capture some general capacities to learn and generalize over patterns.

There's whole range from highly detailed and accurate. Eg. something like OpenWorm and Computational Neuroscience and Multiscale Brain Modeling to your average multi-layer Perceptron in Sci-kit Learn.

Most of the popular networks used today are being optimised for practicality rather than biological accuracy.

Though often researchers start to look at something that was previously ignored by other simplified networks and then decide it might be worth looking at again for its practicality. For example : Spiking Neural Networks, the Next Generation of Machine Learning

Is "spiking" an important idea we've been missing? Well, if it helps solve certain problems, we'll say "yes". If not, we'll leave it to the biologists.