Singular Value Decomposition is a way to break a matrix into three special matrices $$A=U\Sigma V^T$$ SVD basically says, any complicated transformation can be broken down into 3 simple transformations Rotate Stretch and squish Rotate again