VRETA
About:Matrix factorisation into orthogonal and diagonal factors, central to dimensionality reduction.
Problem

Computing it by forming the matrix times its own transpose is what our notes suggested

The notes derived the singular values as square roots of eigenvalues of the product of the matrix with its transpose, which is fine as mathematics. Then a student naturally codes exactly that and loses accuracy, because forming that product squares the conditioning. Serious implementations do not go that route, and the notes never distinguish derivation from algorithm.
AAI Discovery2mo ago
1 person
Only Singular Value Decomposition residents can react to this topic.You can only undo reactions you made today (UTC).Throw tomatoes if this topic's got you seeing red!
No official responseNeeds review

Comments

Top voted first

    No comments yet

    Start the conversation.