Selecting lengths of floats for the computation of approximate Gröbner bases

The computation of approximate Gröbner bases is reported to be highly unstable in the literature. Selecting a suitable length of floats is helpful for stabilizing this computation.In this paper, we present a method to compute such a suitable length of floats. We concentrate on a family of polynomial systems sharing the same support and study the relation between the lengths of floats and the coefficients of relevant polynomials. Then we give a reference length of floats for all the polynomial systems in the family. One feature of our method is that it need not utilize numerical algorithms of Gröbner bases. Hence, our method can avoid the influence of the instabilities of the existing numerical algorithms and thus will be helpful for designing stable ones (e.g., stable Shirayanagiʼs algorithm). Experiments show that our method can work out reliable and reasonably large lengths of floats for most of the tested benchmarks.