Geometric interpretation of Logistic Regression and Linear regression is considered here.

I was going through Logistic regression and Linear regression. In the optimization equation of both following term is used. $$W^{T}.X$$

W is a vector which holds weights of the hyper-plane.

I realized following about the dimensions of the fitted hyper-plane. Want to confirm it.

Let, d = Number of features for both Logistic Regression and Linear Regression.

Logistic Regression case: Fitted hyper-plane is d-dimensional.

Linear Regression case: Fitted hyper-plane is (d + 1) dimensions.


d = 2

feature 1 : weight, feature 2 : height

Logistic Regression: Its a 2 class classification. y : {obsess, normal)

Linear Regression: y: blood pressure (real value)


  • Logistic Regression will fit a 2-D line.
  • Linear Regression will fit a 3-D plane.

Please confirm if this understanding is correct and same happens even in higher dimensions.


Mm, I understand what you want to say, but I think you're slightly wrong with terminology. If you consider a d-dimensional vector space, people usually a hyper-plane (d - 1)-dimensional, because it's a subspace of lower dimensionality. So for your example d = 2 the logistic regression would fit a 1-dimentional line (which would separate 2 classes) and the linear regression would fit a 2-dimentional plane (for every combination of 2 features the prediction would lay on some plane in 3-D space). So the general rule would be logreg -> (d-1)-dimensional hyperplane in d-dimensional vector space, linear regression -> d-dimensional hyperplane in (d+1)-dimensional vector space.


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.