2
$\begingroup$

When reading about machine learning, I've often come across information stating that Bayesian methods in machine learning are effective when you only possess a limited amount of data. As someone who is planning to start learning machine learning, I was curious how it accomplishes this? Lastly, and more specifically, I was especially wondering how and/or if this can be applied to problems such as image classification, when there is a limited amount of data (limited number of images to train the algorithm on)?

Any textbook, research, or other references with regards to the latter would also be greatly appreciated.

Thanks.

$\endgroup$
2
  • 1
    $\begingroup$ As far as I know, a major reason the Bayesian paradigm cam trump the frequentist one in small samples is because of the use of priors. If you use a relatively informative prior, you can get away with less data because information comes from both the data and the prior, as opposed to only coming from the data $\endgroup$
    – Matt Kaye
    Apr 23 at 1:06
  • $\begingroup$ @MattKaye Yes, I agree with this. I didn't really learn this until after I had posted this question. $\endgroup$ Apr 23 at 1:55
1
$\begingroup$

The discussion of Bayesian reasoning and its advantages over frequentist reasoning is very wide. I refer you to the book “Pattern recognition and machine learning” by Cristopher Bishop which is a great book on Bayesian reasoning.

$\endgroup$

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