1
$\begingroup$

Can the features of feature extraction be extracted by developing a feature extraction approach that can be pointed at a large number of feature extraction processes that have successfully extracted features from a variety of classes of things that possess features?

If so, what are those more abstract features likely to be? Can those features of feature extraction processes then be aimed at creating new feature extractors for new classes of things that possess them?

It is requested that answers not digress into general intelligence conjecture but rather keep to the topic of feature extraction abstraction.

$\endgroup$
  • $\begingroup$ i think to do this properly, your best chance is to have a better system for domain alignment... but this question does raise a wonderfil field to research in $\endgroup$ – mshlis Aug 18 at 17:25
  • $\begingroup$ @mshlis, can you suggest in an answer what system might be more aligned to the domain or how to use a somewhat high level adapter pattern to effectively simulate such alignment? Should you, I'm guessing you'd concurrently edify yourselves and others and score a few rep points. I'd certainly enjoy the read. $\endgroup$ – Douglas Daseeco Aug 18 at 19:17
  • $\begingroup$ I would but i recommend defining your question in a more formalized manner, otherwise it seems there may be multiple interpretations as im reading it again $\endgroup$ – mshlis Aug 20 at 0:02
  • $\begingroup$ @mshlis, a mathematical formalization may largely depend on the answer to the question in the original post, "If so, what are those more abstract features likely to be?" Without such a first stab at an answer, an abstraction of the abstraction of all feature extractions would be required to formulate the question, and the answer would likely become a trivially mechanical consequence of that super-abstraction. In such a case, it would be a paper well beyond the depth of even the most interesting SE posts. $\endgroup$ – Douglas Daseeco Sep 3 at 19:21

Your Answer

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

Browse other questions tagged or ask your own question.