To summarize, PCA automatically finds an efficient basis (or feature set) to represent the data. The coefficients $a_k$ change for each data row (unlike in regression where it's fixed for the whole dataset).PCA found the basis $\phi_k$ (while regression needs $\phi_k$ as input).While this is reminiscent of regression, note two important differences: In this viewpoint, what PCA has done is to fit the 10-term formula for $f$ to a data set of 1797 images. I.e., the numbers $a_k = _i$ represent coefficients in a basis expansion with the basis images $\phi_k$ set by $\phi_k = v_k$, and where $x$ represents one of the 64 pixels. Given a one-dimensional data vector $x = ^t$, its mean, or sample mean isĬonsider a multi-dimensional $m \times n$ data array $X$ representing
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |