I am trying to be precise in definitions.

We can solve regression, classification, clusterisation, dimensionality reduction, visualization, feature extraction tasks.

But also there are supervised, unsupervised, ..., tasks of ML.

I see that regression is a part of the supervised learning.

  1. Are there any other particular tasks?
  2. How to make the word "tasks" more clear?

Some of the "tasks" (there are many more, for example anomaly detection; and people just love to invent their own tasks marginally different from before hoping to be the first to spawn a new subdomain...) can both be supervised or unsupervised.


  • PCA is unsupervised dimensionality reduction
  • Fishers' LDA is supervised dimensionality reduction

... and both predate any notion of machine learning, data science or data mining. These things are just rebranded into ML now, squeezed into a very narrow view that never was able to capture the essence of even clustering (beyond k-means, that seems to be the only method machine learners still understand and get taught, as the L2 nature of the method fits their world view well enough). I'd rather avoid making all this "machine learning" - it was not, and it doesn't fit well.

  • $\begingroup$ Thank you for the answer. I have seen your other answers and have an impression that you are the smartest guy in data-science here. Could you please recommend me a multivariate prediction algorithm based to get a forecast of one variable if we have several terabytes of data? $\endgroup$ – user73786 May 15 '19 at 9:14
  • $\begingroup$ Depends on your data. Size is the worst guidance. If it's too large, you can always downsample. $\endgroup$ – Has QUIT--Anony-Mousse May 15 '19 at 18:16

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.