I wish to make a visualization of the separating hyperplane in the SVM algorithm along with my training features. As my feature vectors have 8 dimensions, we cannot directly visualize this. The only potential solution I can think of is to get a projection down to 2 dimensional space.

  1. Will this solution even make sense? I know that SVMs will often make the data linearly separable in higher-dimensions and this may not occur in 2 dimensional space. Would this at least be a decent approximation?

  2. How do people visualize their classification and their training features in practice? Hi


2 Answers 2


Here is a video I found that would be closest to what you are looking for: https://www.youtube.com/watch?v=3liCbRZPrZA

It shows SVM being applied to 3D. It is very difficult to graphically visualize dimensions higher than 3 and would usually cause more confusion.

With higher dimensions all I would use is vectors to and show significant points but this only would make sense if someone understands how it works in lower dimensions


The visualization of high dimensional data is generally difficult. Most of the methods rely on projection the data onto low dimensions (at most three). For example, Principal component analysis (PCA) explores linear combinations of the original data, so that you can pick the first two principal components to visualize. There has been extensive studies on visualization methods, and in the case of SVM, I think this paper, Support vector machines for visualization and dimensionality reduction, is a good start.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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