1: \begin{abstract}

2: We introduce two quantum algorithms for solving structured prediction problems.

3: We show that a stochastic

4: subgradient descent method that uses the quantum minimum finding algorithm and

5: takes its probabilistic failure into account solves the structured prediction problem with a

6: runtime that scales with the square root of the size of the label space, and in

7: $\widetilde O\left(1/\epsilon\right)$ with respect to the precision, $\epsilon$,

8: of the solution.

9: Motivated by robust inference techniques in machine learning,

10: we introduce another quantum algorithm that solves a smooth approximation

11: of the structured prediction problem with a similar quantum speedup in the size

12: of the label space and a similar scaling in the precision parameter. In doing

13: so, we analyze a stochastic gradient algorithm for convex optimization in the

14: presence of an additive error in the calculation of the gradients, and show that

15: its convergence rate does not deteriorate if the additive errors are of the

16: order $O(\sqrt\epsilon)$. This algorithm uses quantum Gibbs sampling at

17: temperature $\Omega (\epsilon)$ as a subroutine. Based on these theoretical

18: observations, we propose a method for using quantum Gibbs samplers to combine

19: feedforward neural networks with probabilistic graphical models for quantum

20: machine learning. Our numerical results using Monte Carlo simulations on an

21: image tagging task demonstrate the benefit of the approach.

22: \end{abstract}

23: