ML Knowledge
Could you explain the concept of error in linear regression and its significance in ensuring precise modeling?
Machine Learning Engineer
Palantir Technologies
Arm
Audible
Okta
Blend
Answers
Anonymous
7 months ago
An error in linear regression is normally distributed. This is to ensure valid statistical inference of the computed coefficients and for hypothesis testing. Intuitively, we want to make sure the error is normally distributed to compensate for both underfitting and overfitting
Anonymous
10 months ago
Linear regression assumes that the noise is gaussian. This would cause error in the approximation if the noise added was not from a symmetric distribution. For example, if the error was drawn from a poisson distribution, then the tail events would overly influence the fit.
Interview question asked to Machine Learning Engineers interviewing at Coursera, Arm, Hootsuite and others: Could you explain the concept of error in linear regression and its significance in ensuring precise modeling?.