Optimally Stopped Optimization

We combine the fields of heuristic optimization and optimal stopping. We propose a strategy for benchmarking randomized optimization algorithms that minimizes the expected total cost for obtaining a good solution with an optimal number of calls to the solver. To do so, rather than letting the objective function alone define a cost to be minimized, we introduce a further cost-per-call of the algorithm. We show that this problem can be formulated using optimal stopping theory. The expected cost is a flexible figure of merit for benchmarking probabilistic solvers that can be computed when the optimal solution is not known, and that avoids the biases and arbitrariness that affect other measures. The optimal stopping formulation of benchmarking directly leads to a real-time, optimal-utilization strategy for probabilistic optimizers with practical impact. We apply our formulation to benchmark simulated annealing on a class of MAX2SAT problems. We also compare the performance of a D-Wave 2X quantum annealer to the HFS solver, a specialized classical heuristic algorithm designed for low tree-width graphs. On a set of frustrated-loop instances with planted solutions defined on up to N = 1098 variables, the D-Wave device is two orders of magnitude faster than the HFS solver, and, modulo known caveats related to suboptimal annealing times, exhibits identical scaling with problem size.

[1]  Mauro Birattari,et al.  How to assess and report the performance of a stochastic algorithm on a benchmark problem: mean or best result on a number of runs? , 2007, Optim. Lett..

[2]  Thomas Bartz-Beielstein,et al.  Experimental Methods for the Analysis of Optimization Algorithms , 2010 .

[3]  Catherine C. McGeoch,et al.  Benchmarking a quantum annealing processor with the time-to-target metric , 2015, 1508.05087.

[4]  M. W. Johnson,et al.  A scalable control system for a superconducting adiabatic quantum optimization processor , 2009, 0907.3757.

[5]  Ryan Babbush,et al.  What is the Computational Value of Finite Range Tunneling , 2015, 1512.02206.

[6]  Óscar Promio Muñoz Quantum Annealing in the transverse Ising Model , 2018 .

[7]  Jason Brownlee,et al.  A note on research methodology and benchmarking optimization algorithms , 2007 .

[8]  John N. Hooker,et al.  Testing heuristics: We have it all wrong , 1995, J. Heuristics.

[9]  Vasil S. Denchev,et al.  Computational multiqubit tunnelling in programmable quantum annealers , 2015, Nature Communications.

[10]  Mark W. Johnson,et al.  Architectural Considerations in the Design of a Superconducting Quantum Annealing Processor , 2014, IEEE Transactions on Applied Superconductivity.

[11]  M. W. Johnson,et al.  Quantum annealing with manufactured spins , 2011, Nature.

[12]  Martin Herdegen Optimal Stopping and Applications Example 2 : American options , 2009 .

[13]  Firas Hamze,et al.  Glassy Chimeras could be blind to quantum speedup: Designing better benchmarks for quantum annealing machines , 2014, 1401.1546.

[14]  R. G. Miller,et al.  Optimal Persistence Policies , 1960 .

[15]  S. Knysh,et al.  Quantum Optimization of Fully-Connected Spin Glasses , 2014, 1406.7553.

[16]  David Siegmund,et al.  Great expectations: The theory of optimal stopping , 1971 .

[17]  Catherine C. McGeoch A Guide to Experimental Algorithmics , 2012 .

[18]  Matthias Troyer,et al.  Optimised simulated annealing for Ising spin glasses , 2014, Comput. Phys. Commun..

[19]  Daniel A. Lidar,et al.  Evidence for quantum annealing with more than one hundred qubits , 2013, Nature Physics.

[20]  Mauricio G. C. Resende,et al.  Designing and reporting on computational experiments with heuristic methods , 1995, J. Heuristics.

[21]  Cedric Yen-Yu Lin,et al.  Different Strategies for Optimization Using the Quantum Adiabatic Algorithm , 2014, 1401.7320.

[22]  M. W. Johnson,et al.  A scalable readout system for a superconducting adiabatic quantum optimization system , 2009, 0905.0891.

[23]  R. Khan,et al.  Sequential Tests of Statistical Hypotheses. , 1972 .

[24]  Andrew D. King,et al.  Performance of a quantum annealer on range-limited constraint satisfaction problems , 2015, ArXiv.

[25]  J. Andel Sequential Analysis , 2022, The SAGE Encyclopedia of Research Design.

[26]  Daniel Nagaj,et al.  Quantum speedup by quantum annealing. , 2012, Physical review letters.

[27]  Samuel J. Gershman,et al.  A Tutorial on Bayesian Nonparametric Models , 2011, 1106.2697.

[28]  Thomas Stützle,et al.  Evaluating Las Vegas Algorithms: Pitfalls and Remedies , 1998, UAI.

[29]  Daniel A. Lidar,et al.  Tunneling and speedup in quantum optimization for permutation-symmetric problems , 2015, 1511.03910.

[30]  Michael H. Goldwasser,et al.  Data Structures, Near Neighbor Searches, and Methodology: Fifth and Sixth DIMACS Implementation Challenges, Proceedings of a DIMACS Workshop, USA, 1999 , 2002, Data Structures, Near Neighbor Searches, and Methodology.

[31]  Nando de Freitas,et al.  From Fields to Trees , 2004, UAI.

[32]  Damian S. Steiger,et al.  Heavy Tails in the Distribution of Time to Solution for Classical and Quantum Annealing. , 2015, Physical review letters.

[33]  M. W. Johnson,et al.  Experimental investigation of an eight-qubit unit cell in a superconducting optimization processor , 2010, 1004.1628.

[34]  Alex Selby Efficient subgraph-based sampling of Ising-type models with frustration , 2014, 1409.3934.

[35]  Hans Werner Gottinger Sequential analysis and optimal stopping , 1975 .

[36]  Celso C. Ribeiro,et al.  TTT plots: a perl program to create time-to-target plots , 2007, Optim. Lett..

[37]  B. Chakrabarti,et al.  Colloquium : Quantum annealing and analog quantum computation , 2008, 0801.2193.

[38]  A. Gilles,et al.  The Art of Computer Systems Performance Analysis (Techniques for Experimental Design, Measurement, Simulation, and Modeling) , 1992 .

[39]  Rosenbaum,et al.  Quantum annealing of a disordered magnet , 1999, Science.

[40]  Daniel A. Lidar,et al.  Defining and detecting quantum speedup , 2014, Science.

[41]  Daniel A. Lidar,et al.  Probing for quantum speedup in spin-glass problems with planted solutions , 2015, 1502.01663.

[42]  G. Stigler The Economics of Information , 1961, Journal of Political Economy.

[43]  C. D. Gelatt,et al.  Optimization by Simulated Annealing , 1983, Science.

[44]  Daniel A. Lidar,et al.  Consistency tests of classical and quantum models for a quantum annealer , 2014, 1403.4228.