Alibaba Cloud Quantum Development Platform: Applications to Quantum Algorithm Design
Weibo:
Abstract:
We report our work on the Alibaba Cloud Quantum Development Platform (AC-QDP). The capability of AC-QDP's computational engine was already reported in \cite{CZH+18, ZHN+19}.In this follow-up article, we demonstrate with figures how AC-QDP helps in testing large-scale quantum algorithms (currently within the QAOA framework). We give new ...More
Code:
Data:
Introduction
- In the past two years significant efforts have been made towards building intermediate-scale quantum devices (NISQs).
- Optimal QAOA sequences for small-cycle free graphs; 2.
- The QAOA consists of two phases; in the first phase, optimal angle sequences (γ, β) are obtained; in the second phase, these parameters are used to generate a certain number of copies of the state |γ, β via a quantum circuit.
Highlights
- In the past two years significant efforts have been made towards building intermediate-scale quantum devices (NISQs)
- The capability of Alibaba Cloud Quantum Development Platform (AC-QDP)’s computational engine was already reported in [8, 20]. In this follow-up article, we focus on showing how AC-QDP helps in testing of large-scale quantum algorithms, currently within the Quantum Approximate Optimization Algorithm (QAOA) framework
- The QAOA consists of two phases; in the first phase, optimal angle sequences (γ, β) are obtained; in the second phase, these parameters are used to generate a certain number of copies of the state |γ, β via a quantum circuit
- To compare our QAOA simulator with existing quantum simulation package equipped with QAOA functionalities, we choose MAX-CUT problems on a random regular graph and compare the time spent for a single energy function query
- Cirq and Qiskit use the state-vector evolution, whereas qTorch is based on tensor-network simulation with a featured QAOA implementation
- Each software is given the same random instance of a regular graph, and is asked to perform 5 times the energy query to obtain an average time for a single query
Results
- In the case that only the QAOA energy function value is needed, tensor-network based approach allows them to handle much bigger instances than the authors could ever hope for with the state-vector approach.
- Cirq and Qiskit use the state-vector evolution, whereas qTorch is based on tensor-network simulation with a featured QAOA implementation.
- Fig. 1 shows the performance of AC-QDP compared to other softwares, for degree 3 cases: Figure 1: Benchmarking results for random 3-regular graphs.
- The authors apply our QAOA software on a family of instances called small-cycle-free regular graphs.
- All kregular graphs with the same number of vertices and a girth at least 2p + 2 yeild identical QAOA energy functions Fp, and the simulation of QAOA on such instances can be readily reduced to the single lightcone on the unique tree-like subgraph, parametrized by the degree d and the depth p.
- In [11], 1-level QAOA is compared with classical one-step local algorithm on triangle-free regular graphs.
- This suggests that the optimal angle sequences for the tree-like case is usually a good starting point for the QAOA when dealing with large regular graphs.
- It is proposed in [16] that the angle sequences for small-cycle-free instances be used directly to completely get rid of the optimization phase, which could be chanllenging on noisy quantum devices.
- Thanks to the tensor-network-based implementation, the software is able to efficiently query small-cycle-free instances of large degree d and number of layers p, albeit the number of vertices involved in the tree-like subgraphs might be large.
Conclusion
- Even when the conjecture holds, it is not sufficient to put the graph isomorphism problem in quantum polynomial time, since the difference between the energy values can be exponentially small.
- The authors have extensively tested the QAOA-based algorithm described above with the Alibaba Cloud Quantum Development Platform, and the authors could separate all non-isomorphic three regular graphs up-to size 18, all strongly regular graphs up-to size 26 and Miyazaki graphs of size 20.
- The depth one QAOA values for 3-regular graphs with respect to the CUT energy function is solely the function of the degree sequence, the number of nodes and the number of triangles of the graph.
Summary
- In the past two years significant efforts have been made towards building intermediate-scale quantum devices (NISQs).
- Optimal QAOA sequences for small-cycle free graphs; 2.
- The QAOA consists of two phases; in the first phase, optimal angle sequences (γ, β) are obtained; in the second phase, these parameters are used to generate a certain number of copies of the state |γ, β via a quantum circuit.
- In the case that only the QAOA energy function value is needed, tensor-network based approach allows them to handle much bigger instances than the authors could ever hope for with the state-vector approach.
- Cirq and Qiskit use the state-vector evolution, whereas qTorch is based on tensor-network simulation with a featured QAOA implementation.
- Fig. 1 shows the performance of AC-QDP compared to other softwares, for degree 3 cases: Figure 1: Benchmarking results for random 3-regular graphs.
- The authors apply our QAOA software on a family of instances called small-cycle-free regular graphs.
- All kregular graphs with the same number of vertices and a girth at least 2p + 2 yeild identical QAOA energy functions Fp, and the simulation of QAOA on such instances can be readily reduced to the single lightcone on the unique tree-like subgraph, parametrized by the degree d and the depth p.
- In [11], 1-level QAOA is compared with classical one-step local algorithm on triangle-free regular graphs.
- This suggests that the optimal angle sequences for the tree-like case is usually a good starting point for the QAOA when dealing with large regular graphs.
- It is proposed in [16] that the angle sequences for small-cycle-free instances be used directly to completely get rid of the optimization phase, which could be chanllenging on noisy quantum devices.
- Thanks to the tensor-network-based implementation, the software is able to efficiently query small-cycle-free instances of large degree d and number of layers p, albeit the number of vertices involved in the tree-like subgraphs might be large.
- Even when the conjecture holds, it is not sufficient to put the graph isomorphism problem in quantum polynomial time, since the difference between the energy values can be exponentially small.
- The authors have extensively tested the QAOA-based algorithm described above with the Alibaba Cloud Quantum Development Platform, and the authors could separate all non-isomorphic three regular graphs up-to size 18, all strongly regular graphs up-to size 26 and Miyazaki graphs of size 20.
- The depth one QAOA values for 3-regular graphs with respect to the CUT energy function is solely the function of the degree sequence, the number of nodes and the number of triangles of the graph.
Tables
- Table1: Time for a single query in seconds, for regular graphs with n = 1000 vertices
- Table2: Comparison of different optimizers on small-cycle-free regular graphs, together with the best values obtained. Cases where the experiment was infeasible is indicated by ‘-’. Best function values found are highlighted. The number of vertices and the averaged time per query in ACQDP are also listed
- Table3: Best angle sequences found by classical optimizers in AC-QDP
Funding
- We would like to thank our colleagues from various teams in Alibaba Cloud Intelligence supporting us in the numerical experiments presented in this paper
Study subjects and analysis
cases: 3
The regular graphs are chosen randomly, running through 10, 20, 30, 50, 100 and 1000 vertices with degree 3, 4 and 5. Fig. 1 shows the performance of AC-QDP compared to other softwares, for degree 3 cases: Figure 1: Benchmarking results for random 3-regular graphs. For each number of vertices, 5 random instances are drawn, and QAOA of depth 1 through 4 are simulated
Reference
- Cirq: A python framework for creating, editing, and invoking Noisy Intermediate Scale Quantum (NISQ) circuits, 2019.
- Héctor Abraham et al. Qiskit: An open-source framework for quantum computing, 2019.
- László Babai. Graph isomorphism in quasipolynomial time [extended abstract]. In Proceedings of the 48th Annual ACM SIGACT Symposium on Theory of Computing, STOC 2016, Cambridge, MA, USA, June 18-21, 2016, pages 684–697, 2016.
- László Babai, Paul Erdös, and Stanley M. Selkow. Random graph isomorphism. SIAM J. Comput., 9(3):628–635, 1980.
- Étienne Bamas and Louis Esperet. Local approximation of the maximum cut in regular graphs. arXiv preprint arXiv:1902.04899, 2019.
- S. Boixo, S. V. Isakov, V. N. Smelyanskiy, and H. Neven. Simulation of low-depth quantum circuits as complex undirected graphical models. ArXiv e-prints, December 2017.
- Kamil Bradler, Shmuel Friedland, Josh Izaac, Nathan Killoran, and Daiqin Su. Graph isomorphism and gaussian boson sampling, 2018.
- Jianxin Chen, Fang Zhang, Cupjin Huang, Michael Newman, and Yaoyun Shi. Classical simulation of intermediate-size quantum circuits. arXiv preprint arXiv:1805.01450, 2018.
- David Emms, Simone Severini, Richard C. Wilson, and Edwin R. Hancock. Coined quantum walks lift the cospectrality of graphs and trees. Pattern Recognition, 42(9):1988–2002, 2009.
- Edward Farhi, Jeffrey Goldstone, and Sam Gutmann. A quantum approximate optimization algorithm. arXiv preprint arXiv:1411.4028, 2014.
- Matthew B Hastings. Classical and quantum bounded depth approximation algorithms. arXiv preprint arXiv:1905.07047, 2019.
- Eugene M. Luks. Isomorphism of graphs of bounded valence can be tested in polynomial time. J. Comput. Syst. Sci., 25(1):42–65, 1982.
- A Mahasinghe, J A Izaac, J B Wang, and J K Wijerathna. Phase-modified CTQW unable to distinguish strongly regular graphs efficiently. Journal of Physics A: Mathematical and Theoretical, 48(26):265301, jun 2015.
- Igor L. Markov and Yaoyun Shi. Simulating quantum computation by contracting tensor networks. SIAM Journal on Computing, 38(3):963–981, 2008.
- E. Schuyler Fried, N. P. D. Sawaya, Y. Cao, I. D. Kivlichan, J. Romero, and A. Aspuru-Guzik. qTorch: The Quantum Tensor Contraction Handler. ArXiv e-prints, September 2017.
- Michael Streif and Martin Leib. Training the Quantum Approximate Optimization Algorithm without access to a quantum processing unit. arXiv preprint arXiv:1908.08862, 2019.
- Mario Szegedy. Graph structure discovery with QAOA. in preparation, 2019.
- H. Wang, J. Wu, X. Yang, and X. Yi. A graph isomorphism algorithm using signatures computed via quantum walk search model. Journal of Physics A Mathematical General, 48(11):115302, March 2015.
- Zhihui Wang, Stuart Hadfield, Zhang Jiang, and Eleanor G Rieffel. Quantum approximate optimization algorithm for MaxCut: A fermionic view. Physical Review A, 97(2):022304, 2018.
- Fang Zhang, Cupjin Huang, Michael Newman, Junjie Cai, Huanjun Yu, Zhengxiong Tian, Bo Yuan, Haihong Xu, Junyin Wu, Xun Gao, et al. Alibaba Cloud Quantum Development Kit: Large-Scale Classical Simulation of Quantum Circuits. arXiv preprint arXiv:1907.11217, 2019.
- Leo Zhou, Sheng-Tao Wang, Soonwon Choi, Hannes Pichler, and Mikhail D Lukin. Quantum Approximate Optimization Algorithm: performance, mechanism, and implementation on near-term devices. arXiv preprint arXiv:1812.01041, 2018.
Full Text
Tags
Comments