12
$\begingroup$

In what situations should we use t-SNE (apart from data visualization)?

T-SNE is used for dimensionality reduction. The answer to this question suggests that t-SNE should be used only for visualization and that we should not use it for clustering. Then what is the good use for t-SNE?

$\endgroup$
  • 6
    $\begingroup$ The standard advice is to NOT use tsne for clustering, because the clusters are so dependent on the perplexity. It is supposed to only be used for "visualization". But that isn't very clear to me, as one immediately looks for (and sees) clusters when looking at a tsne plot. Therefore your question is a good one: what is tsne good for? $\endgroup$ – generic_user Apr 4 '17 at 1:11
  • 2
    $\begingroup$ See the answer to this question that I asked: stats.stackexchange.com/questions/263539/… $\endgroup$ – generic_user Apr 4 '17 at 1:12
  • $\begingroup$ like the @generic_user said, i want to know the benefit of t-sne, beside it's visualization. $\endgroup$ – wolfe Apr 7 '17 at 16:13
  • $\begingroup$ I don't know why this was closed as a duplicate. OP is asking what are the good uses of t-sne apart from visualization. The linked thread is all about clustering. But there might be other uses. $\endgroup$ – amoeba Apr 27 '17 at 11:04
  • $\begingroup$ Related: stats.stackexchange.com/questions/132639/… $\endgroup$ – Amelio Vazquez-Reina Apr 27 '17 at 12:42
4
$\begingroup$

The answer to this question suggests that t-SNE should be used only for visualization and that we should not use it for clustering. Then what is the good use for t-SNE?

I don't agree with this conclusion. There is no reason to assume that t-SNE is any worse universally than any other clustering algorithm. Every clustering algorithm makes assumptions about the structure of the data, and they can be expected to perform differently depending on the underlying distribution and end use of the reduced dimensionality.

t-SNE like many unsupervised learning algorithms often provide a means to an end, e.g. obtaining early insight on whether or not the data is separable, testing that it has some identifiable structure, and inspecting the nature of this structure. One does not need visualization of the t-SNE output to start answering some of these questions. Other applications of lower dimensional embeddings include building features for classification or getting rid of multi-collinearity to improve the performance of prediction methods.

| cite | improve this answer | |
$\endgroup$

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.