6
$\begingroup$

I am producing a research statement to be sent to a statistics department and I was trying to avoid the term Machine learning in favour of the more friendly one of Statistical learning. Probably I could not avoid such use.

$\endgroup$
3
  • 7
    $\begingroup$ Yes, it's completely reasonable to call ANNs statistical learning. You may enjoy Leo Breiman's discussion paper Statistical Modeling: The Two Cultures. $\endgroup$ – Arya McCarthy May 14 at 14:47
  • 3
    $\begingroup$ people in stats departments usually know what neural networks are, if they are not interested in that topic, they won't be interested in it even if you call stat. learning if they are interested in that topic, they will also be interested in it if you call it machine learning. $\endgroup$ – rep_ho May 14 at 14:50
  • $\begingroup$ What about "statistical inference"? $\endgroup$ – mhdadk May 14 at 22:38
8
$\begingroup$

The classic The Elements of Statistical Learning handbook by Hastie et al discusses neural networks among other algorithms, so it needs to be a “statistical learning” algorithm.

Depending whom you’d ask, neural networks are either statistics, statistical learning, pattern recognition, machine learning, deep learning, or artificial intelligence. There’s no single, agreed category used by everybody to describe them.

$\endgroup$
3
  • 1
    $\begingroup$ I think the answer is implied in this statement, but it isn't quite explicit. It may be worth adding another sentence ;-). $\endgroup$ – gung - Reinstate Monica May 14 at 19:09
  • $\begingroup$ I know the textbook and that was one of the reasons suggesting me to do so. Yet others disagree. $\endgroup$ – Mr Frog May 14 at 19:15
  • $\begingroup$ @MrFrog it’s either statistics, statistical learning, pattern recognition, machine learning, deep learning, or artificial intelligence. Each one of those has it’s fans and enemies. There’s no single term that everybody uses. $\endgroup$ – Tim May 14 at 19:35
3
$\begingroup$

That's a political question, not a statistical one :-)

Historically, statistics and machine learning were two distinct communities, with little interaction. ANNs were developed by the machine learning community. Today, the lines might be somewhat blurred, with some statisticians counting ANNs to statistics, while some machine learners count logistic regression and even linear regression to machine learning. Needless to say, some members of the opposite camp beg to differ.

So, there is no simple answer to your question. Calling ANNs statistical method might be seen as justified by some and objected to by others.

$\endgroup$

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.