1. What about uj which corresponds to eigenvalue 0? (Is it still linear combination of phi?)

2. Can we assume the dot product is the standard one (same as multiply a row vector and a column one)

3. Can the coefficients calculation use the uj itself? R we know the uj and its eigenvalue?

In other words, the question asks for "getting uj coefficients, assume u know uj", or "describe the full kernel pca, and show how to use the coefficients instead of the vectors themself"?

Thnx!