Computing group rank with limited nondeterminism

An efficient algorithm computing the rank of a group (that is, the size of a minimum generating subset) benefits mathematicians, who use numerical algebra systems for research, cryptographers, who rely on algebraic systems for proofs of security, and theoretical computer scientists, who seek to understand which problems can be solved in a particular model of computation. Before now, the best algorithm for computing the rank of a group required a polylogarithmic amount of space, which induces a superpolynomial (hence, inefficient) algorithm. We reduced the best upper bound on the complexity of the group rank problem and provide a theoretically efficient algorithm for it. This paper proves that with very short certificates of correctness, the group rank problem can be verified by highly restricted models of computation. We prove that the problem of deciding whether the rank of a finite group, given as a multiplication table, is smaller than a specified number is decidable not only by a circuit of depth $O(\log \log n)$ augmented with $O(\log^2 n)$ nondeterministic bits, but also by a Turing machine using $O(\log n)$ space and $O(\log^2 n)$ bits of nondeterminism. These models of computation are extremely limited in computational power, and hence can be simulated by a deterministic Turing machine in polynomial time. Using limited nondeterminism and restrictive models of computation as verifiers may be useful in examining other algebraic problems.