On the Quantum versus Classical Learnability of Discrete Distributions

Here we study the comparative power of classical and quantum learners for generative modelling within the Probably Approximately Correct (PAC) framework. More specifically we consider the following task: Given samples from some unknown discrete probability distribution, output with high probability an efficient algorithm for generating new samples from a good approximation of the original distribution. Our primary result is the explicit construction of a class of discrete probability distributions which, under the decisional Diffie-Hellman assumption, is provably not efficiently PAC learnable by a classical generative modelling algorithm, but for which we construct an efficient quantum learner. This class of distributions therefore provides a concrete example of a generative modelling problem for which quantum learners exhibit a provable advantage over classical learning algorithms. In addition, we discuss techniques for proving classical generative modelling hardness results, as well as the relationship between the PAC learnability of Boolean functions and the PAC learnability of discrete probability distributions.

[1]  Nader H. Bshouty,et al.  Learning DNF over the uniform distribution using a quantum example oracle , 1995, COLT '95.

[2]  Ievgeniia Oshurko.  Quantum Machine Learning , 2020, Quantum Computing.

[3]  Dan Boneh,et al.  The Decision Diffie-Hellman Problem , 1998, ANTS.

[4]  Lei Wang,et al.  Differentiable Learning of Quantum Circuit Born Machine , 2018, Physical Review A.

[5]  Akimasa Miyake,et al.  Quantum supremacy in constant-time measurement-based computation: A unified architecture for sampling and verification , 2017, 1703.11002.

[6]  Yehuda Lindell,et al.  Introduction to Modern Cryptography (Chapman & Hall/Crc Cryptography and Network Security Series) , 2007 .

[7]  Hans-J. Briegel,et al.  Machine learning \& artificial intelligence in the quantum domain , 2017, ArXiv.

[8]  Alon Orlitsky,et al.  On Learning Distributions from their Samples , 2015, COLT.

[9]  Reza Rezaeian Farashahi,et al.  Efficient Pseudorandom Generators Based on the DDH Assumption , 2007, Public Key Cryptography.

[10]  Ilias Diakonikolas,et al.  Learning Structured Distributions , 2016, Handbook of Big Data.

[11]  Scott Aaronson,et al.  The computational complexity of linear optics , 2010, STOC '11.

[12]  Nathan Killoran,et al.  Quantum generative adversarial networks , 2018, Physical Review A.

[13]  Silvio Micali,et al.  How to construct random functions , 1986, JACM.

[14]  Andrew W. Cross,et al.  Demonstration of quantum advantage in machine learning , 2015, npj Quantum Information.

[15]  J. Eisert,et al.  Architectures for quantum simulation showing a quantum speedup , 2017, 1703.00466.

[16]  Max Welling,et al.  Auto-Encoding Variational Bayes , 2013, ICLR.

[17]  Manuel Blum,et al.  How to generate cryptographically strong sequences of pseudo random bits , 1982, 23rd Annual Symposium on Foundations of Computer Science (sfcs 1982).

[18]  Andrew W. Cross,et al.  Quantum learning robust against noise , 2014, 1407.5088.

[19]  Jack Hidary,et al.  Quantum Hamiltonian-Based Models and the Variational Quantum Thermalizer Algorithm , 2019, ArXiv.

[20]  Yoshua Bengio,et al.  Generative Adversarial Nets , 2014, NIPS.

[21]  Mark Zhandry,et al.  How to Construct Quantum Random Functions , 2012, 2012 IEEE 53rd Annual Symposium on Foundations of Computer Science.

[22]  Michael Kharitonov,et al.  Cryptographic lower bounds for learnability of Boolean functions on the uniform distribution , 1992, COLT '92.

[23]  Michele Mosca,et al.  Exact quantum Fourier transforms and discrete logarithm algorithms , 2003 .

[24]  Travis S. Humble,et al.  Quantum supremacy using a programmable superconducting processor , 2019, Nature.

[25]  Vikraman Arvind,et al.  The Complexity of Exactly Learning Algebraic Concepts. (Extended Abstract) , 1996, ALT.

[26]  Ashley Montanaro,et al.  Average-case complexity versus approximate simulation of commuting quantum computations , 2015, Physical review letters.

[27]  N. V. VinodchandranInstitute.  The Complexity of Exactly Learning Algebraic Concepts , 1996 .

[28]  Ronitt Rubinfeld,et al.  On the learnability of discrete distributions , 1994, STOC '94.

[29]  Ivan Kobyzev,et al.  Normalizing Flows: An Introduction and Review of Current Methods , 2020, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[30]  Shouvanik Chakrabarti,et al.  Quantum Wasserstein Generative Adversarial Networks , 2019, NeurIPS.

[31]  Leslie G. Valiant,et al.  Cryptographic Limitations on Learning Boolean Formulae and Finite Automata , 1993, Machine Learning: From Theory to Applications.

[32]  Oded Goldreich,et al.  Foundations of Cryptography: Volume 1, Basic Tools , 2001 .

[33]  Karsten A. Verbeurgt.  Learning DNF under the uniform distribution in quasi-polynomial time , 1990, COLT '90.

[34]  Umesh V. Vazirani,et al.  An Introduction to Computational Learning Theory , 1994 .

[35]  Ronald de Wolf,et al.  Guest Column: A Survey of Quantum Learning Theory , 2017, SIGA.

[36]  Leonard Pitt,et al.  On the necessity of Occam algorithms , 1990, STOC '90.

[37]  Krzysztof Pietrzak,et al.  Cryptography from Learning Parity with Noise , 2012, SOFSEM.

[38]  Leslie G. Valiant,et al.  A theory of the learnable , 1984, STOC '84.

[39]  Jens Eisert,et al.  Sample Complexity of Device-Independently Certified "Quantum Supremacy". , 2018, Physical review letters.

[40]  Michael Kharitonov,et al.  Cryptographic hardness of distribution-specific learning , 1993, STOC.

[41]  David Haussler,et al.  Learnability and the Vapnik-Chervonenkis dimension , 1989, JACM.

[42]  Rocco A. Servedio,et al.  Equivalences and Separations Between Quantum and Classical Learnability , 2004, SIAM J. Comput..

[43]  Andrea Rocchetto,et al.  Statistical limits of supervised quantum learning , 2020, Physical Review A.

[44]  Shai Ben-David,et al.  Understanding Machine Learning: From Theory to Algorithms , 2014 .

[45]  Shu-Hao Wu,et al.  Quantum generative adversarial learning in a superconducting quantum circuit , 2018, Science Advances.

[46]  Elham Kashefi,et al.  The Born supremacy: quantum advantage and training of an Ising Born machine , 2019, npj Quantum Information.

[47]  Moni Naor,et al.  Number-theoretic constructions of efficient pseudo-random functions , 2004, JACM.

[48]  Alon Rosen,et al.  Pseudorandom Functions: Three Decades Later , 2017, Tutorials on the Foundations of Cryptography.

[49]  Masahito Hayashi,et al.  Verifying commuting quantum computations via fidelity estimation of weighted graph states , 2019, New Journal of Physics.

[51]  Seth Lloyd,et al.  Quantum Generative Adversarial Learning. , 2018, Physical review letters.

[52]  David Haussler,et al.  Equivalence of models for polynomial learnability , 1988, COLT '88.

[53]  L. Duan,et al.  Quantum Supremacy for Simulating a Translation-Invariant Ising Spin Model. , 2016, Physical review letters.