Subtropical Real Root Finding
Thomas Sturm · 2015
We describe a new incomplete but terminating heuristic method for real root finding for large multivariate polynomials. We take an abstract view of the polynomial as the set of exponent vectors associated with sign information on the coefficients. Then we employ linear programming to heuristically find roots. There is a specialized variant for roots with exclusively positive coordinates, which is of considerable interest for applications in chemistry and systems biology. An implementation of our method combining the computer algebra system Reduce with the linear programming solver Gurobi has been successfully applied to input data originating from established mathematical models used in these areas. We have solved several hundred problems with up to more than 800,000 monomials in up to 10 variables with degrees up to 12. Our method has failed due to its incompleteness in only 10 percent of the cases.