3
$\begingroup$

I was watching a documentary on Netflix about AlphaGo, and at one point (~1:10:16 from the end), one of the programmers uses the term "heavy node," which I assume has to do with neural networks. I did a little bit of research but couldn't find anything on what that term means. The closest I could get was this wikipedia page on Heavy path decomposition: https://en.wikipedia.org/wiki/Heavy_path_decomposition, which seemed like it could be somewhat related, but I wasn't sure how exactly. Has anyone heard of this term being used? Does anyone know what it means?

For context, in the documentary the line is that if it (the network/player) creates something new not in the heavy node, then they don't see it.

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ I'm not aware of what a heavy node is. I remember watching this movie and that term didn't catch my attention. If you're interested in some of the inner workings of AlphaGo, you could check out their latest paper - mooc360.com/agz_unformatted_nature.pdf (note this is AlphaGo Zero - the successor to AlphaGo featured in Netflix) $\endgroup$
    – Zakk Diaz
    Commented Jan 15, 2018 at 19:45

1 Answer 1

Reset to default
4
$\begingroup$

Yes, firstly you are correct to acknowledge the term node with reference to the neural network.

A neural network is made to mimic the working of a human brain hence the term neural. It comprises of several layers and each layer comprises of several nodes. A network can be a deep neural network, signifying the presence of numerous layers, in which the output of the one, serves as an input for the other. Further a node in a particular layer has some weights and biases attached with it that are responsible for the computation and producing output.

To your context, a heavy node may refer to a node that holds importance for the network and has got significant weight and bias attached. Henceforth as per the statement if the network does make any significant change to the node with importance in terms of weights and biases, it is noticed else not.

Improve this answer
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .