Regarding an adaptive algorithm for testing multivariate linear dependence (Q616299)

The author examines the complexity of his adaptive algorithm [Linear Algebra Appl. 408, 151--160 (2005; Zbl 1075.32001)] for testing linear dependence of \(N\) multivariate functions within some differential field of \(m\) variables and compare it with the complexity of generalized Wronskian techniques. It is found that the number of derivaive rows that must be calculated by this algorithm can be bounded by \(N\) plus a term that is sublinear with respect to \(N\), thus of the order \(N\) as a whole. A combinatorial argument yields a sharp bound on the size of the marginal set \({\mathcal B}_Y=\{\alpha\in{\mathbb N}^m\setminus Y\, |\, Y\cup\{a\} \text{ is Young-like}\}\) of a Young-like set \(Y\) in term of the size of \(Y\).