1
$\begingroup$

I am reading an article about graph neural network and it is mentioned:

In this step, we extract all newly update hidden states and create a final feature vector describing the whole graph. This feature vector can be then used as input to a standard machine learning model.

What does it mean that this feature vector can be used as an input to standard machine learning model? Isnt machine learning all about obtaining the features in the first place? And what does it imply that feature vector would be input to machine learning model? Should it be part of the ML model rather an input?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

There's quite a lot of confusion here:

  • The word "features" is a bit vague but it usually refers to the structured information provided as input to a ML system.
  • No, features are not a part a ML model. They are the input used to train a model.
  • No, ML is not "all about obtaining features". Obtaining the features is just the stage of obtaining the information in a usable way, so it's as if you said that "learning" is all about "obtaining a book".

In this sentence the author means there are (at least) two stages to this ML process: the first step trains a neural network model, then the values of the hidden states of the NN are collected in order to provide them to a second model which uses them as input features.

Improve this answer
$\endgroup$
4
  • $\begingroup$ Ok so this is what I understood, the first step trains a neural network model (to obtain features in the form of hidden states)? to serve as an input features to second model, so in this context is the second model also a neural network? $\endgroup$
    – user0193
    Commented Feb 6, 2022 at 7:31
  • 1
    $\begingroup$ 1) yes. 2) the second model doesn't have to be a neural network, in fact they say a "standard" ML model so I think they mean a traditional (non-neural) model. But in general it could be any kind. $\endgroup$
    – Erwan
    Commented Feb 6, 2022 at 13:13
  • $\begingroup$ +1 it makes soo much sense now! So a "standard" ML model could be anything like Gaussian process or non-linear regression etc. right? $\endgroup$
    – user0193
    Commented Feb 6, 2022 at 19:44
  • $\begingroup$ can you please also take a look at this $\endgroup$
    – user0193
    Commented Apr 4, 2022 at 21:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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