A single $T$-gate makes distribution learning hard

Abstract

The task of learning a probability distribution from samples is ubiquitous across the natural sciences. The output distributions of local quantum circuits form a particularly interesting class of distributions, of key importance both to quantum advantage proposals and a variety of quantum machine learning algorithms. In this work, we provide an extensive characterization of the learnability of the output distributions of local quantum circuits. Our first result yields insight into the relationship between the efficient learnability and the efficient simulatability of these distributions. Specifically, we prove that the density modelling problem associated with Clifford circuits can be efficiently solved, while for depth $d=n^{\Omega (1)}$ circuits the injection of a single $T$-gate into the circuit renders this problem hard. This result shows that efficient simulatability does not imply efficient learnability. Our second set of results provides insight into the potential and limitations of quantum generative modelling algorithms. We first show that the generative modelling problem associated with depth $d=n^{\Omega (1)}$ local quantum circuits is hard for any learning algorithm, classical or quantum. As a consequence, one cannot use a quantum algorithm to gain a practical advantage for this task. We then show that, for a wide variety of the most practically relevant learning algorithms -- including hybrid-quantum classical algorithms -- even the generative modelling problem associated with depth $d=\omega (\log (n))$ Clifford circuits is hard. This result places limitations on the applicability of near-term hybrid quantum-classical generative modelling algorithms.

5+12 pages, 3 figures

References in corpus (4)

• Classical simulation of commuting quantum computations implies collapse of the polynomial hierarchy
• Beyond Barren Plateaus: Quantum Variational Algorithms Are Swamped With Traps
• Fast simulation of stabilizer circuits using a graph state representation
• The Learnability of Quantum States

Cited by in corpus (22)

• Stabilizer entropy dynamics after a quantum quench
• Learning efficient decoders for quasi-chaotic quantum scramblers
• Learning quantum states and unitaries of bounded gate complexity
• Stabilizer entropy in non-integrable quantum evolutions
• Exponentially tighter bounds on limitations of quantum error mitigation
• Improved Stabilizer Estimation via Bell Difference Sampling
• Spectral Properties Versus Magic Generation in $T$-doped Random Clifford Circuits
• Efficient learning of $t$-doped stabilizer states with single-copy measurements
• Random unitaries, Robustness, and Complexity of Entanglement
• Efficient Learning of Quantum States Prepared With Few Non-Clifford Gates
• Simple Hamiltonian dynamics is a powerful quantum processing resource
• Anticoncentration and State Design of Doped Real Clifford Circuits and Tensor Networks
• On the average-case complexity of learning output distributions of quantum circuits
• Quantum Local Differential Privacy and Quantum Statistical Query Model
• Learning unitaries with quantum statistical queries
• Learning Quantum Processes with Quantum Statistical Queries
• PAC-learning of free-fermionic states is NP-hard
• Agnostic Process Tomography
• On the Hardness of Measuring Magic
• Anticoncentration in Clifford Circuits and Beyond: From Random Tensor Networks to Pseudo-Magic States
• Optimal quantum reservoir learning in proximity to universality
• Gottesman-Knill Limit on One-way Communication Complexity: Tracing the Quantum Advantage down to Magic Resources