Effective algorithms for computing triangular operator in Schubert calculus (Q2258133)

The triangular operator plays an important role in the study of Schubert calculus on Schubert calculus, which concerns intersection numbers of Schubert varieties. Since the Bott-Samelson manifolds provide smooth resolution of Schubert varieties by the classical work of \textit{R. Bott} and \textit{H. Samelson} [Proc. Natl. Acad. Sci. USA 41, 490--493 (1955; Zbl 0064.25903)], the operator \(T_{A_n}\) has played a fundamental role in performing Schubert calculus in homogeneous spaces. For example, in terms of the triangle operator, one can derive formulae for evaluating the degrees of Schubert varieties, computing the Schubert structure constants on homogeneous spaces \(G/H\), determining the Steenrod operations on Schubert varieties, and finding presentations for the integral cohomology ring of \(G/H\), where \(G\) is a compact connected Lie group with \(H\subset G\) being the centralizer of a one-parameter subgroup. We know that the integral cohomology ring \(H^\ast (E_8 /T)\) of the homogeneous spaces \(E_8 / T\) is generated by Schubert generators in degrees \(n \leq 15\), subject to recursive relations in degrees \(n\leq 30\), where \(T\subset G = E_8\) is a maximal torus. Some applications to the geometry and the topology of Schubert varieties rely on computational results via the triangular operator. There have been programs that resort to certain built-in functions in \texttt{Mathematica}. However, they are very consuming both in time and memory when the degree \(n\) tends to large in practice. The reason is that, these functions are designed on the recursive relations, which are suitable for large-scale computing. Furthermore, these built-in functions are not suitable for parallel computing. To overcome these drawbacks, the authors present two parallel algorithms implemented by C++ which evaluate the triangular operator in a much more efficient manner. The new algorithms are based on a newly-designed iterative relation instead of the original recursive one, which accelerates the computation significantly. In this work, they first introduce three preliminary algorithms (Preliminary Algorithms A, B, and C, abbreviated by PAA, PAB, and PAC), which pave the way for the two main parallel algorithms (Algorithms 1 and 2). The first two preliminary algorithms (PAA and PAB) are designed by using the iterative relation forwardly, and the corresponding computational complexities are approximately proportionate to the Catalan number and the Motzkin number, respectively. While the third preliminary algorithm (PAC) is designed by using the iterative relation backwardly, and the computational complexity is about central binomial coefficient. Starting from PAC, we propose Algorithm \(1\), a parallel solver which makes the triangular operator problem solvable when \(n\leq 26\). The benchmark case of the triangular operator is when \(n = 30\), which is sufficient for performing Schubert calculus in the most difficult case. Finally, an enhanced parallel solver (Algorithm 2) based on PAC is presented for solving the challenging case \(n= 30\) efficiently.