1204.2821 (Vadim N. Smelyanskiy et al.)

A Near-Term Quantum Computing Approach for Hard Computational Problems
in Space Exploration    [PDF]

Vadim N. Smelyanskiy, Eleanor G. Rieffel, Sergey I. Knysh, Colin P. Williams, Mark W. Johnson, Murray C. Thom, William G. Macready, Kristen L. Pudenz

In this article, we show how to map a sampling of the hardest artificial intelligence problems in space exploration onto equivalent Ising models that then can be attacked using quantum annealing implemented in D-Wave machine. We overview the existing results as well as propose new Ising model implementations for quantum annealing. We review supervised and unsupervised learning algorithms for classification and clustering with applications to feature identification and anomaly detection. We introduce algorithms for data fusion and image matching for remote sensing applications. We overview planning problems for space exploration mission applications and algorithms for diagnostics and recovery with applications to deep space missions. We describe combinatorial optimization algorithms for task assignment in the context of autonomous unmanned exploration. Finally, we discuss the ways to circumvent the limitation of the Ising mapping using a "blackbox" approach based on ideas from probabilistic computing. In this article we describe the architecture of the D-Wave One machine and report its benchmarks. Results on random ensemble of problems in the range of up to 96 qubits show improved scaling for median core quantum annealing time compared with classical algorithms; whether this scaling persists for larger problem sizes is an open question. We also review previous results of D-Wave One benchmarking studies for solving binary classification problems with a quantum boosting algorithm which is shown to outperform AdaBoost. We review quantum algorithms for structured learning for multi-label classification and introduce a hybrid classical/quantum approach for learning the weights. Results of D-Wave One benchmarking studies for learning structured labels on four different data sets show a better performance compared with an independent Support Vector Machine approach with linear kernel.

View original: http://arxiv.org/abs/1204.2821