1: \begin{abstract}

2: 

3: A challenging problem in computational mathematics is to

4: compute roots of a high-degree univariate random polynomial.

5: We combine an efficient multiprecision implementation 

6: for solving high-degree random polynomials with two 

7: certification methods, namely Smale's $\alpha$-theory and one 

8: based on Gerschgorin's theorem, for showing 

9: that a given numerical approximation is in the quadratic 

10: convergence region of Newton's method of some exact solution. 

11: With this combination, we can {\it certifiably count}

12: the number of real roots of random polynomials.  We quantify the difference

13: between the two certification procedures and list the salient features of both of them. 

14: After benchmarking on random polynomials where the coefficients are drawn from the Gaussian distribution, 

15: we obtain novel experimental results for the~Cauchy~distribution~case.

16: \end{abstract}

17: