Please let me ask a question.

As I understood from the very clear explanation from Professor Andrew, the PCA allows to dramatically reduce he number of parameters characterizing the object. The new, reduced set o f parameters, consists of linear combinations or the original parameters. So, these linear combinations are not more physically meaningful, though they nicely describe our object, right?

Will be grateful to you for the answer.

1 Like

Yes you can say that! If a point of information is 10 dimensional than you can reduce it to 3 dimensions, lets say by using projections and in 3 dimensions the object is well represented but it has lost some information in the other dimensions, but maybe you don’t even need those other dimensions (depending on you application!)

Thank you for the answer.

Yes, but the 3 remained dimensions are not the original data, but some linear combinations of them, aren’t they?

1 Like

You are correct.