1
$\begingroup$

Everywhere in the theory of neural networks, authors saying that idea came about by observing the work of the human brain. But I can not believe in that. I guess, everything is much simpler and neural networks is specific functions of math's series. Proof of this is the existence of the Weierstrass theorem and Taylor series, which says that every function can be approximated by certain polynomials. Am I right?

$\endgroup$
  • 2
    $\begingroup$ "I guess everything is much simpler and neural networks is specific functions of math's series" It is unclear what you are asking. Also note that the Weierstrass theorem does not approximate using Taylor series -- for one thing, it applies to arbitrary continuous functions, even nowhere differentiable ones. $\endgroup$ – Chill2Macht Mar 27 '17 at 11:08
1
$\begingroup$

Neural network idea can be very loosely explained as following:

a. Using enough hidden layers can lead to the representation of every function that exists (again, this is loosely and is far from practical)

b. Since now you can represent every function in the world with parameters (weights) now use stochastic gradient decent on the weights (this is the back propagation in super short) to get to the optimal classifier (or regressor)

$\endgroup$
  • $\begingroup$ But how scientists look at neural networks? $\endgroup$ – Dmitry Nalyvaiko Mar 27 '17 at 11:18
  • $\begingroup$ which scientists? can you be more specific? $\endgroup$ – Cherny Mar 27 '17 at 11:19
  • $\begingroup$ mathematics and those who study neural networks from a scientific point of view $\endgroup$ – Dmitry Nalyvaiko Mar 27 '17 at 11:21
  • $\begingroup$ So I don't understand what you mean when you say scientific (really sorry but it's not very clear what you mean). But about mathematics: The whole process of machine learning is looking for the best function in a subspace of all the functions. Neural network can represent a very considerable subspace, then you need to get optimal function from that subspace. And if you need more specification about the connection to brain: Neurons have a connection to a lot of other neurons, and the connection can be change over time. And not only that but the neuron depends non-linearly on it's input $\endgroup$ – Cherny Mar 27 '17 at 11:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.