Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
194 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
46 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Barren Plateaus in Variational Quantum Computing (2405.00781v2)

Published 1 May 2024 in quant-ph, cs.LG, and stat.ML

Abstract: Variational quantum computing offers a flexible computational paradigm with applications in diverse areas. However, a key obstacle to realizing their potential is the Barren Plateau (BP) phenomenon. When a model exhibits a BP, its parameter optimization landscape becomes exponentially flat and featureless as the problem size increases. Importantly, all the moving pieces of an algorithm -- choices of ansatz, initial state, observable, loss function and hardware noise -- can lead to BPs when ill-suited. Due to the significant impact of BPs on trainability, researchers have dedicated considerable effort to develop theoretical and heuristic methods to understand and mitigate their effects. As a result, the study of BPs has become a thriving area of research, influencing and cross-fertilizing other fields such as quantum optimal control, tensor networks, and learning theory. This article provides a comprehensive review of the current understanding of the BP phenomenon.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (108)
  1. M. Schuld, I. Sinayskiy, and F. Petruccione, An introduction to quantum machine learning, Contemporary Physics 56, 172 (2015).
  2. L. Bittel and M. Kliesch, Training variational quantum algorithms is NP-hard, Phys. Rev. Lett. 127, 120502 (2021).
  3. E. R. Anschuetz and B. T. Kiani, Beyond barren plateaus: Quantum variational algorithms are swamped with traps, Nature Communications 13, 7760 (2022a).
  4. E. R. Anschuetz, Critical points in quantum generative models, International Conference on Learning Representations  (2022).
  5. P. Bermejo, B. Aizpurua, and R. Orús, Improving gradient methods via coordinate transformations: Applications to quantum machine learning, Physical Review Research 6, 023069 (2024).
  6. S. Sim, P. D. Johnson, and A. Aspuru-Guzik, Expressibility and entangling capability of parameterized quantum circuits for hybrid quantum-classical algorithms, Advanced Quantum Technologies 2, 1900070 (2019).
  7. Q. Miao and T. Barthel, Equivalence of cost concentration and gradient vanishing for quantum circuits: an elementary proof in the riemannian formulation, arXiv preprint arXiv:2402.07883  (2024).
  8. M. Cerezo and P. J. Coles, Higher order derivatives of quantum neural networks with barren plateaus, Quantum Science and Technology 6, 035006 (2021).
  9. J. L. Cybulski and T. Nguyen, Impact of barren plateaus countermeasures on the quantum neural network capacity to learn, Quantum Information Processing 22, 442 (2023).
  10. A. Pérez-Salinas, H. Wang, and X. Bonet-Monroig, Analyzing variational quantum landscapes with information content, npj Quantum Information 10, 27 (2024).
  11. S. Okumura and M. Ohzeki, Fourier coefficient of parameterized quantum circuits and barren plateau problem, arXiv preprint arXiv:2309.06740  (2023).
  12. N. A. Nemkov, E. O. Kiktenko, and A. K. Fedorov, Fourier expansion in variational quantum algorithms, Phys. Rev. A 108, 032406 (2023).
  13. C. O. Marrero, M. Kieferová, and N. Wiebe, Entanglement-induced barren plateaus, PRX Quantum 2, 040316 (2021).
  14. D. García-Martín, M. Larocca, and M. Cerezo, Deep quantum neural networks form gaussian processes, arXiv preprint arXiv:2305.09957  (2023).
  15. A. A. Mele, Introduction to haar measure tools in quantum information: A beginner’s tutorial, arXiv preprint arXiv:2307.08956  (2023).
  16. A. Uvarov and J. D. Biamonte, On barren plateaus and cost function locality in variational quantum algorithms, Journal of Physics A: Mathematical and Theoretical 54, 245301 (2021).
  17. T. Barthel and Q. Miao, Absence of barren plateaus and scaling of gradients in the energy optimization of isometric tensor network states, arXiv preprint arXiv:2304.00161  (2023).
  18. Q. Miao and T. Barthel, Isometric tensor network optimization for extensive hamiltonians is free of barren plateaus, arXiv preprint arXiv:2304.14320  (2023).
  19. A. Letcher, S. Woerner, and C. Zoufal, Tight and efficient gradient bounds for parameterized quantum circuits, arXiv preprint arXiv:2309.12681  (2023).
  20. H.-K. Zhang, S. Liu, and S.-X. Zhang, Absence of barren plateaus in finite local-depth circuits with long-range entanglement, Physical Review Letters 132, 150603 (2024a).
  21. J. Napp, Quantifying the barren plateau phenomenon for a model of unstructured variational ansätze, arXiv preprint arXiv:2203.06174  (2022).
  22. C. Zhao and X.-S. Gao, Analyzing the barren plateau phenomenon in training quantum neural networks with the ZX-calculus, Quantum 5, 466 (2021).
  23. C.-Y. Park and N. Killoran, Hamiltonian variational ansatz without barren plateaus, Quantum 8, 1239 (2024).
  24. C.-Y. Park, M. Kang, and J. Huh, Hardware-efficient ansatz without barren plateaus in any depth, arXiv preprint arXiv:2403.04844  (2024).
  25. M. Schumann, F. K. Wilhelm, and A. Ciani, Emergence of noise-induced barren plateaus in arbitrary layered noise models, arXiv preprint arXiv:2310.08405  (2023).
  26. P. Singkanipa and D. A. Lidar, Beyond unital noise in variational quantum algorithms: noise-induced barren plateaus and fixed points, arXiv preprint arXiv:2402.08721  (2024).
  27. R. Zeier and T. Schulte-Herbrüggen, Symmetry principles in quantum systems theory, Journal of mathematical physics 52, 113510 (2011).
  28. M. J. Bremner, C. Mora, and A. Winter, Are random pure states useful for quantum computation?, Physical review letters 102, 190502 (2009).
  29. D. Gross, S. T. Flammia, and J. Eisert, Most quantum states are too entangled to be useful as computational resources, Physical review letters 102, 190501 (2009).
  30. M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information (Cambridge University Press, Cambridge, 2000).
  31. J. Horgan, Scott aaronson answers every ridiculously big question i throw at him, Scientific American. https://blogs. scientificamerican. com/cross-check/scott-aaronson-answers-every-ridiculously-big-question-i-throw-at-him 21 (2016).
  32. Y. Liu, S. Arunachalam, and K. Temme, A rigorous and robust quantum speed-up in supervised machine learning, Nature Physics , 1 (2021).
  33. J. Jäger and R. V. Krems, Universal expressiveness of variational quantum classifiers and quantum kernels for support vector machines, Nature Communications 14, 576 (2023).
  34. L. Friedrich and J. Maziero, Quantum neural network cost function concentration dependency on the parametrization expressivity, Scientific Reports 13, 9978 (2023).
  35. T. Haug, K. Bharti, and M. S. Kim, Capacity and Quantum Geometry of Parametrized Quantum Circuits, PRX Quantum 2, 040309 (2021).
  36. Technically speaking, group modules are irreducible representations of the dynamical Lie group over the vector space ℬ⁢(ℋ0)ℬsubscriptℋ0\mathcal{B}(\mathcal{H}_{0})caligraphic_B ( caligraphic_H start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ).
  37. A. Uvarov, J. D. Biamonte, and D. Yudin, Variational quantum eigensolver for frustrated quantum systems, Physical Review B 102, 075104 (2020).
  38. M. Kashif and S. Al-Kuwari, The impact of cost function globality and locality in hybrid quantum neural networks on nisq devices, Machine Learning: Science and Technology 4, 015004 (2023a).
  39. O. Ogunkoya, K. Morris, and D. M. Kürkçüoglu, Investigating parameter trainability in the snap-displacement protocol of a qudit system, arXiv preprint arXiv:2309.14942  (2023).
  40. B. Zhang and Q. Zhuang, Energy-dependent barren plateau in bosonic variational quantum circuits, arXiv preprint arXiv:2305.01799  (2023).
  41. R. Shaydulin and S. M. Wild, Importance of kernel bandwidth in quantum machine learning, Physical Review A 106, 042407 (2022).
  42. S. Das, S. Martina, and F. Caruso, The role of data embedding in equivariant quantum convolutional neural networks, arXiv preprint arXiv:2312.13250  (2023).
  43. D. Stilck França and R. Garcia-Patron, Limitations of optimization algorithms on noisy quantum devices, Nature Physics 17, 1221 (2021).
  44. M. M. Wilde, Quantum information theory (Cambridge University Press, 2013).
  45. D. García-Martín, M. Larocca, and M. Cerezo, Effects of noise on the overparametrization of quantum neural networks, Phys. Rev. Res. 6, 013295 (2024).
  46. D. Wecker, M. B. Hastings, and M. Troyer, Progress towards practical quantum variational algorithms, Physical Review A 92, 042303 (2015).
  47. E. Farhi, J. Goldstone, and S. Gutmann, A quantum approximate optimization algorithm, arXiv preprint arXiv:1411.4028  (2014).
  48. A. Bärtschi and S. Eidenbenz, Grover mixers for qaoa: Shifting complexity from mixer design to state preparation, in 2020 IEEE International Conference on Quantum Computing and Engineering (QCE) (IEEE, 2020) pp. 72–82.
  49. B. Zhang, A. Sone, and Q. Zhuang, Quantum computational phase transition in combinatorial problems, npj Quantum Information 8, 1 (2022b).
  50. A. G. Taube and R. J. Bartlett, New perspectives on unitary coupled-cluster theory, International journal of quantum chemistry 106, 3393 (2006).
  51. R. Mao, G. Tian, and X. Sun, Barren plateaus of alternated disentangled ucc ansatzs (2023).
  52. S. C. Marshall, C. Gyurik, and V. Dunjko, High dimensional quantum machine learning with small quantum computers, Quantum 7, 1078 (2023).
  53. T. Volkoff and P. J. Coles, Large gradients via correlation in random parameterized quantum circuits, Quantum Science and Technology 6, 025008 (2021).
  54. S. Kazi, M. Larocca, and M. Cerezo, On the universality of snsubscript𝑠𝑛s_{n}italic_s start_POSTSUBSCRIPT italic_n end_POSTSUBSCRIPT-equivariant k𝑘kitalic_k-body gates, arXiv preprint arXiv:2303.00728  (2023).
  55. R. Jozsa and A. Miyake, Matchgates and classical simulation of quantum circuits, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 464, 3089 (2008).
  56. F. De Melo, P. Ćwikliński, and B. M. Terhal, The power of noisy fermionic quantum computation, New Journal of Physics 15, 013015 (2013).
  57. T. J. Volkoff, Efficient trainability of linear optical modules in quantum optical neural networks, Journal of Russian Laser Research 42, 250 (2021).
  58. T. Volkoff, Z. Holmes, and A. Sornborger, Universal compiling and (no-)free-lunch theorems for continuous-variable quantum learning, PRX Quantum 2, 040327 (2021).
  59. M. Arjovsky, A. Shah, and Y. Bengio, Unitary evolution recurrent neural networks, in International conference on machine learning (PMLR, 2016) pp. 1120–1128.
  60. A. Kulshrestha and I. Safro, Beinit: Avoiding barren plateaus in variational quantum algorithms, in 2022 IEEE International Conference on Quantum Computing and Engineering (QCE) (IEEE, 2022) pp. 197–203.
  61. A. Rad, A. Seif, and N. M. Linke, Surviving the barren plateau in variational quantum circuits with bayesian learning initialization, arXiv preprint arXiv:2203.02464  (2022).
  62. T. Haug and M. Kim, Optimal training of variational quantum algorithms without barren plateaus, arXiv preprint arXiv:2104.14543  (2021).
  63. X. Shi and Y. Shang, Avoiding barren plateaus via gaussian mixture model, arXiv preprint arXiv:2402.13501  (2024).
  64. L. Friedrich and J. Maziero, Avoiding barren plateaus with classical deep neural networks, Physical Review A 106, 042433 (2022).
  65. G. Marin-Sanchez, J. Gonzalez-Conde, and M. Sanz, Quantum algorithms for approximate function loading, Physical Review Research 5, 033114 (2023).
  66. A. Cervera-Lierta, J. S. Kottmann, and A. Aspuru-Guzik, The meta-variational quantum eigensolver (meta-vqe): Learning energy profiles of parameterized Hamiltonians for quantum simulation, PRX Quantum 2, 020329 (2021).
  67. J. Wurtz and D. Lykov, Fixed-angle conjectures for the quantum approximate optimization algorithm on regular maxcut graphs, Physical Review A 104, 052419 (2021).
  68. S. Boulebnane and A. Montanaro, Predicting parameters for the quantum approximate optimization algorithm for max-cut from the infinite-size limit, arXiv preprint arXiv:2110.10685  (2021).
  69. E. Campos, A. Nasrallah, and J. Biamonte, Abrupt transitions in variational quantum circuit training, Physical Review A 103, 032607 (2021b).
  70. G. Acampora, A. Chiatto, and A. Vitiello, A comparison of evolutionary algorithms for training variational quantum classifiers, in 2023 IEEE Congress on Evolutionary Computation (CEC) (IEEE, 2023) pp. 1–8.
  71. B. Koczor and S. C. Benjamin, Quantum natural gradient generalized to noisy and nonunitary circuits, Physical Review A 106, 062416 (2022).
  72. K. Temme, S. Bravyi, and J. M. Gambetta, Error mitigation for short-depth quantum circuits, Physical review letters 119, 180509 (2017).
  73. Y. Li and S. C. Benjamin, Efficient variational quantum simulator incorporating active error minimization, Phys. Rev. X 7, 021050 (2017).
  74. K. Wang, Y.-A. Chen, and X. Wang, Mitigating quantum errors via truncated neumann series, Science China Information Sciences 66, 180508 (2023b).
  75. R. Takagi, H. Tajima, and M. Gu, Universal sampling lower bounds for quantum error mitigation, Physical Review Letters 131, 210602 (2023).
  76. J.-G. Liu and L. Wang, Differentiable learning of quantum circuit born machines, Phys. Rev. A 98, 062324 (2018).
  77. C. Zoufal, Generative quantum machine learning, arXiv preprint arXiv:2111.12738  (2021).
  78. S. Lloyd and C. Weedbrook, Quantum generative adversarial learning, Physical Review Letters 121, 040502 (2018).
  79. L. Coopmans and M. Benedetti, On the sample complexity of quantum boltzmann machine learning, arXiv preprint arXiv:2306.14969  (2023).
  80. M. Kieferova, O. M. Carlos, and N. Wiebe, Quantum generative training using rényi divergences, arXiv preprint arXiv:2106.09567  (2021).
  81. J. Kübler, S. Buchholz, and B. Schölkopf, The inductive bias of quantum kernels, Advances in Neural Information Processing Systems 34, 12661 (2021).
  82. Y. Suzuki and M. Li, Effect of alternating layered ansatzes on trainability of projected quantum kernel, arXiv preprint arXiv:2310.00361  (2023).
  83. Y. Suzuki, H. Kawaguchi, and N. Yamamoto, Quantum fisher kernel for mitigating the vanishing similarity issue, arXiv preprint arXiv:2210.16581  (2022).
  84. M. Schuld and F. Petruccione, Supervised learning with quantum computers, Vol. 17 (Springer, 2018).
  85. M. Schuld, Supervised quantum machine learning models are kernel methods, arXiv preprint arXiv:2101.11020  (2021).
  86. B. T. Kiani, S. Lloyd, and R. Maity, Learning unitaries by gradient descent, arXiv preprint arXiv:2001.11897  (2020).
  87. X. Ge, R.-B. Wu, and H. Rabitz, The optimization landscape of hybrid quantum–classical algorithms: From quantum control to NISQ applications, Annual Reviews in Control https://doi.org/10.1016/j.arcontrol.2022.06.001 (2022).
  88. L. Broers and L. Mathey, Mitigated barren plateaus in the time-nonlocal optimization of analog quantum-algorithm protocols, Physical Review Research 6, 013076 (2024).
  89. H.-X. Tao, J. Hu, and R.-B. Wu, Unleashing the expressive power of pulse-based quantum neural networks, arXiv preprint arXiv:2402.02880  (2024).
  90. E. C. Martín, K. Plekhanov, and M. Lubasch, Barren plateaus in quantum tensor network optimization, Quantum 7, 974 (2023).
  91. V. Feldman, Statistical query learning, in Encyclopedia of Algorithms, edited by M.-Y. Kao (Springer New York, New York, NY, 2016) pp. 2090–2095.
  92. A. Angrisani, Learning unitaries with quantum statistical queries, arXiv preprint arXiv:2310.02254  (2023).
  93. C. Wadhwa and M. Doosti, Learning quantum processes with quantum statistical queries, arXiv preprint arXiv:2310.02075  (2023).
  94. A. Nietner, Unifying (quantum) statistical and parametrized (quantum) algorithms, arXiv preprint arXiv:2310.17716  (2023).
  95. L. Yang and N. Engelhardt, The complexity of learning (pseudo) random dynamics of black holes and other chaotic systems, arXiv preprint arXiv:2302.11013  (2023).
  96. E. R. Anschuetz and B. T. Kiani, Quantum variational algorithms are swamped with traps, Nature Communications 13, 7760 (2022b).
  97. X. You and X. Wu, Exponentially many local minima in quantum neural networks, in International Conference on Machine Learning (PMLR, 2021) pp. 12144–12155.
  98. A. Tikku and I. H. Kim, Circuit depth versus energy in topologically ordered systems, arXiv preprint arXiv:2210.06796  (2022).
  99. K. Bharti and T. Haug, Iterative quantum-assisted eigensolver, Physical Review A 104, L050401 (2021a).
  100. K. Bharti and T. Haug, Quantum-assisted simulator, Physical Review A 104, 042418 (2021b).
  101. C. Gyurik, R. Molteni, and V. Dunjko, Limitations of measure-first protocols in quantum machine learning, arXiv preprint arXiv:2311.12618  (2023).
  102. X. Glorot and Y. Bengio, Understanding the difficulty of training deep feedforward neural networks, in Proceedings of the thirteenth international conference on artificial intelligence and statistics (JMLR Workshop and Conference Proceedings, 2010) pp. 249–256.
  103. S. Ioffe and C. Szegedy, Batch normalization: Accelerating deep network training by reducing internal covariate shift, in International conference on machine learning (pmlr, 2015) pp. 448–456.
  104. J. L. Ba, J. R. Kiros, and G. E. Hinton, Layer normalization, arXiv preprint arXiv:1607.06450  (2016).
  105. V. Nair and G. E. Hinton, Rectified linear units improve restricted boltzmann machines, in Proceedings of the 27th international conference on machine learning (ICML-10) (2010) pp. 807–814.
  106. S. Hochreiter, The vanishing gradient problem during learning recurrent neural nets and problem solutions, International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems 6, 107 (1998).
  107. S. Hochreiter and J. Schmidhuber, Long short-term memory, Neural computation 9, 1735 (1997).
  108. L. Baird and A. Moore, Gradient descent for general reinforcement learning, Advances in neural information processing systems 11 (1998).
Citations (64)
View on Semantic Scholar

Summary

  • The paper finds that barren plateaus, arising from high-dimensional parameter spaces and circuit expressivity, result in vanishing gradients during quantum circuit training.
  • The study shows that noise and decoherence further exacerbate these effects, significantly hindering optimization in variational quantum computing.
  • The paper explores mitigation strategies such as layerwise training, smart initialization, and adaptive circuit design to enhance the scalability of VQCs.

Understanding Barren Plateaus in Variational Quantum Computing

What Are Barren Plateaus?

Barren Plateaus (BPs) in the context of variational quantum computing (VQC) refer to regions within the parameter landscape of quantum circuits where the gradient of the objective function essentially vanishes, making optimization exceedingly difficult. As quantum circuit depth increases or as the number of qubits gets larger, landscapes tend to flatten out exponentially, leading to these barren plateaus.

Why Do They Matter?

Experts in quantum computing are enthusiastic about VQCs because of their potential to solve complex problems more efficiently than classical computers, particularly in fields like chemistry and cryptography. BPs are a significant hurdle in realizing this potential because they make training quantum models challenging, resulting in slower learning and poorer performance.

How Do Barren Plateaus Occur?

Several factors contribute to the emergence of barren plateaus:

  • Increase in System Size: As the number of qubits increases, the dimensionality of the Hilbert space (quantum state space) they span grows exponentially. High-dimensional spaces often exhibit properties where randomly chosen vectors (states) are nearly orthogonal, leading to very small gradients.
  • High Expressivity of Quantum Circuits: When quantum circuits are capable of expressing a very large set of unitary transformations (making them highly expressive), they explore the Hilbert space very thoroughly. Ironically, this thorough exploration can lead to averaging effects where the gradients of the objective function average out to zero, leading to barren plateaus.
  • Noise and Decoherence: Practical quantum devices are noisy. Research indicates that certain types of noise and decoherence can exacerbate the flatness of the optimization landscape, deepening barren plateaus.

Practical Consequences

If not properly addressed, barren plateaus can severely limit the scalability of VQCs. This limitation is not just about longer training times but also concerns the feasibility of obtaining meaningful quantum computational advantages over classical approaches.

Strategies to Mitigate Barren Plateaus

Several techniques have been proposed and tested to mitigate or avoid barren plateaus:

  • Layerwise Training: Training shallow layers of the quantum circuit first and gradually increasing the depth.
  • Smart Initialization: Choosing initial parameters carefully, potentially using classical pre-training methods to find a good starting point.
  • Use of Local Costs: Designing cost functions that depend only on a subset of qubits or employ local observables.
  • Adaptive Circuit Design: Dynamically adjusting the circuit's architecture during training to navigate away from flat regions.

Future Directions

Understanding barren plateaus is an ongoing area of research. Future work includes developing more effective strategies for initialization and adaptive circuit design, exploring the role of symmetries and other inductive biases in preventing barren plateaus, and quantifying the impact of noise on training dynamics. Further theoretical insights into the structure of high-dimensional quantum landscapes will also be crucial for devising new algorithms that are inherently resistant to barren plateaus.

Conclusion

Barren plateaus represent a fundamental challenge in the field of quantum computing, particularly affecting the trainability of variational quantum algorithms. By continuing to address these challenges through innovative strategies and deeper understanding, researchers aim to unlock the full potential of quantum computing technologies. As this field evolves, overcoming barren plateaus will be critical for achieving practical quantum advantage.

File Document Download Save Streamline Icon: https://streamlinehq.com PDF File Document Download Save Streamline Icon: https://streamlinehq.com Markdown