Research On Classical And Quantum Simulation Of Boson-sampling

Quantum computing theoretically expresses its great power over classical computing,and Boson-Sampling process is probable to be the first model to verify this comparison practically.Quantum algorithms offer exponential speed-up over classical ones,with Shor's factorization algorithm as the outstanding example.However,building a“universal”quantum computer that runs this algorithm is desperately hard,which requires the implementation of the quantum logical gate that conducts precise manipulation on quantum states.A device contains quantum bits of large quantity is beyond the capability of current technology.Fortunately,the new raised model,Boson-Sampling in2010 brings a new approach verify the ability of quantum computing.Boson-sampling encodes the information used in calculation with the Fock state of the bosons(photons,for example),and can present great power without logical gate.Therefore,boson-sampling is thought to be the paragon to show the great power of quantum computing,and attracts great attention from the globe.“The least scale for quantum computing overcoming classical computing”,“The construction of the evolution matrix in Boson-sampling”,etc.are key problems in the study of Boson-sampling.Concentrating on the former,we studied the classical simulation on the fastest classical computer——Milkyway-2,trying to get the performance limitation.For the latter,we studied the construction of network in Boson-sampling,and conducted boson-sampling experiment.The work and contribution of this paper are listed as follows:1.We implemented the optimal classical algorithm and completed the test in massive parallelism on Milkyway-2.The test result indicates that using 13,000 nodes that contain312,000 cores,and solved the problem in which the matrix scales up to 48.2.Based on the features of Milkyway-2,the heterogeneous parallelism of CPU and co-processors of MIC(Many Integrated Cores)architecture was implemented.We reorganized the program using vector instructions for the execution of the 512-bit vector processing unit in MIC.The test result shows that the code running on MIC obtains the speedup of 1.7 over that running on CPU with 24 computing cores.3.We implemented the algorithm for constructing the optical network corresponding to a given matrix,and verified the availability of this algorithm with a simple case.For the unknown parameters in the optical network building by wave guide,we simulated the process of measuring these parameters using random tests,and analyzed the effects of the system errors on the measurement.The simulation shows that this program is capable to factorize arbitrary matrix,and the error in system affects less to the measurement.4.We implemented the two-photon boson-sampling process with linear optical experiment using discrete optical elements,and analyzed the obtained data.The test shows that identical photons are prepared and two-photon boson-sampling process is conducted.