Los Alamos National Laboratory LA-UR-20-28072 The Quantum Alternating Operator Ansatz on Maximum k-Vertex Cover you joint work with Jeremy Cook and Stephan Eidenbenz Andreas BΓ€rtschi nt CCS-3 Information Sciences baertschi@lanl.gov wo IEEE International Conference on Quantum Computing and Engineering (QCE20) October 12 β 16, 2020 Managed by Triad National Security, LLC for the U.S. Department of Energy's NNSA
Los Alamos National Laboratory QAOA and k-Vertex Cover October 12, 2020 | 2
Los Alamos National Laboratory Quantum Alternating Operator Ansatz (QAOA) QAOA is a heuristic for constraint combinatorial optimization. Given a problem over inputs π¦ β πΈ β 0,1 ! with objective function π π¦ βΆ πΈ β β , prepare a state πΏ = π "#$ ! % " π "#& ! % # β― π "#$ $ % " π "#& $ % # π β πΎ, β from which one would like to sample good solutions with high probability. QAOA is specified by β’ an initial state π in a feasible subspace given by the domain πΈ , β’ a phase separating cost Hamiltonian πΌ ' , diagonal in the computational basis: πΌ ' π¦ = π(π¦) π¦ , β’ a mixing Hamiltonian πΌ ( preserving and interfering solutions in πΈ , β’ π rounds with individual angle parameters πΎ ) , β¦ , πΎ * , πΏ ) , β¦ , πΏ * . October 12, 2020 | 3
Los Alamos National Laboratory Maximum k-Vertex Cover Maximize the number of Edges covered by π out of all π Vertices of an input Graph (here π = 4 , π = 9 ) . π¦ . = 0 β’ Greedy : 10 edges covered β’ Maximum : 11 edges covered π¦ ) = 1 β’ In general (for non-constant π ): NP-hard to approximate to 1 β π , UG-hard to approximate to 0.944, classical polynomial-time 0.92-approximation Single Equality Constraint β πβπΎ π π = π Simple objective function October 12, 2020 | 4
Los Alamos National Laboratory Maximum k-Vertex Cover Maximize the number of Edges covered by π out of all π Vertices of an input Graph (here π = 4 , π = 9 ) . π(π¦) = ? ππ(π¦ / , π¦ 1 ) π¦ . = 0 /,1 β2 π¦ ) = 1 " 1 β π ! = ? 1 β (1 β π¦ / )(1 β π¦ 1 ) π¦ ! = 2 /,1 β2 πΌ ' = 1 3 β π / 3 β π 1 3 π 1 3 4 ? 3 π½π β π / /,1 β2 Single Equality Constraint β πβπΎ π π = π Simple objective function October 12, 2020 | 5
Los Alamos National Laboratory Experiments We compared different QAOA approaches for Maximum π -Vertex Cover ! on random ErdΕs-RΓ©nyi π» !,* graphs on π = 7,8,9,10 vertices with π = β . β and edge probability π = 0.5 : Initial States Mixers Angle Selection Strategies Dicke States Ring Mixer Monte Carlo Random States Complete Graph Mixer Basin Hopping Interpolation Equal Superposition of all Hamming-Weight π Basis States, Dicke States : Randomly chosen Hamming-Weight π Basis State. Random State : Ring Mixer : Hamming-weight preserving XY-interactions on a Ring, Complete Mixer : Hamming-weight preserving XY-interactions on a Clique. October 12, 2020 | 7
Los Alamos National Laboratory Agenda Initial State and Mixer Choice β’ Dicke States vs. Random States β’ Ring Mixer vs. Complete Graph Mixer Angle Selection Strategies β’ Angle Correlations β’ Strategy evaluations Summary October 12, 2020 | 8
Los Alamos National Laboratory Initial State and Mixer Choice October 12, 2020 | 9
Los Alamos National Laboratory Hamming Weight π Subspace Problem is constraint to solutions π¦ of Hamming Weight π . We need to start in this subspace and stay in it: Initial State Mixer Choice for π "ππΈπ° π΅ Random Basis State |π¦β© of 0 0 0 0 β’ 0 0 1 0 : = Hamming Weight x = π , 5 + π ) : π . 5 π . π ) 0 1 0 0 or 0 0 0 0 β’ Dicke State : π #>) : ! 5 π #>) 5 πΌ ;#!< = β #6= π # + π # β’ 1 ! = πΈ 4 ? |π¦β© π 5 + π # : π : 5 π πΌ ?@#A/B = β #CD π # 5 64 β’ π D D October 12, 2020 | 10
Los Alamos National Laboratory Complete Graph vs. Ring Mixer and Dicke vs. k-state Complete Graph Mixer Ring Mixer β’ Complete Graph mixer finds better solutions at lower round counts β’ Dicke starting state outperforms randomly selected k-state starting state October 12, 2020 | 11
Los Alamos National Laboratory Complete Graph Mixer outperforms Ring Mixer Approximation ratio of the complete mixer π π³ compared to the approximation ratio of the ring mixer π πΊ for several rounds: β’ Complete Graph Mixer consistently outperforms Ring Mixer (by a ratio > 1 even w/ confidence intervals) β’ Ratio decreases for higher rounds, when both Mixers are close to the optimum. (Plot data taken over 100 graphs of size 7) October 12, 2020 | 12
Los Alamos National Laboratory Angle Selection Strategies October 12, 2020 | 13
Los Alamos National Laboratory Monte Carlo Angle Sampling does not work well β’ Monte Carlo Angle Sampling: take p random angles for gamma and beta β’ Plot (right) shows a declining approximation ratio if a total of 1000 sample angles are distributed across π rounds (Data on 10 random graphs; Complete Graph mixer; Dicke states) β’ Plot (right) shows number of Monte Carlo samples required to get an increase in best approximation ratios found ( π wrt π β 1 ) Exponential scaling in π makes Monte Carlo β’ not seem promising as angle selection strategy October 12, 2020 | 14
Los Alamos National Laboratory Angle Selection: Are angles good for many graphs? Complete Graph Mixer Ring Mixer Heat plots show expectation values for 1-round QAOA averaged over 100 graphs. Γ Hope for generally good values. October 12, 2020 | 15
Los Alamos National Laboratory Are good angles correlated across rounds and graphs? Beta Gamma Plot shows best πΎ and πΏ values for π = 6 -round QAOAs of typical graphs. Red line: Average over all tested graphs (graph size 7 β 10) . October 12, 2020 | 16
Los Alamos National Laboratory Good angles for increasing rounds p ? π = 6 π = 7 π = 5 πΎ Ξ³ Correlation holds across different values of the number of rounds π . Γ Interpolation strategy for angle selection October 12, 2020 | 17
Los Alamos National Laboratory Comparison of Angle Selection Strategies Strategies β’ Monte Carlo : βrandom anglesβ β’ Basin Hopping : βdetect and escape local minimaβ β’ Interpolation : βlearn angles from other graphsβ β’ Optimal : βfull angle exploration at 0.01π resolutionβ Experimental Setup β’ Same number of samples for all three strategies and for all p Findings: only interpolation manages to increase performance up to 4 rounds, Basin hopping and Monte Carlo do not profit from > π round. October 12, 2020 | 18
Los Alamos National Laboratory Summary October 12, 2020 | 19
Los Alamos National Laboratory Conclusion k -Vertex Cover QAOA Main Findings β’ Dicke States outperform Random π -states as initial state β’ Complete Graph Mixer outperforms Ring Mixer β’ Interpolation method finds angles that are good across graphs and rounds Additional Findings β’ Monte Carlo angle sampling works poorly β’ Ring mixer not periodic, no circuit implementation of Complete Graph Mixer Future work & Open questions β’ Do results hold for larger graphs? Conjecture: performance gaps increase β’ How well does the Grover Mixer perform? (see other talk in this session) β’ Do findings generalize to other optimization problems? October 12, 2020 | 20
Recommend
More recommend