Exponential Separations Between Learning With and Without Quantum Memory

  title={Exponential Separations Between Learning With and Without Quantum Memory},
  author={Sitan Chen and Jordan S. Cotler and Hsin-Yuan Huang and Jerry Zheng Li},
  journal={2021 IEEE 62nd Annual Symposium on Foundations of Computer Science (FOCS)},
We study the power of quantum memory for learning properties of quantum systems and dynamics, which is of great importance in physics and chemistry. Many state-of-the-art learning algorithms require access to an additional external quantum memory. While such a quantum memory is not required a priori, in many cases, algorithms that do not utilize quantum memory require much more data than those which do. We show that this trade-off is inherent in a wide range of learning problems. Our results… 

Figures and Tables from this paper

Quantum advantage in learning from experiments
This research presents a probabilistic model of the black hole that combines quantum entanglement, superposition, and superposition to describe the fabric of space-time.
Scalable measures of magic for quantum computers
Magic characterizes the degree of non-stabilizerness of quantum states. It is a crucial resource for quantum computing and a necessary condition for quantum advantage. However, quantifying magic
Tight Bounds for Quantum State Certification with Incoherent Measurements
This work focuses on algorithms which use incoherent measurements, i.e. which only measure one copy of ρ at a time, and can be implemented without persistent quantum memory and thus represent a large class of protocols that can be run on current or near-term devices.
Tight Bounds for State Tomography with Incoherent Measurements
This work fully resolves the question of whether or not this rate of coherent measurements is tight, by showing that any protocol using incoherent measurements, even if they are chosen adaptively, requires Ω(d/ε) copies, matching the upper bound of [KRT17].
Distributed Quantum inner product estimation
The sample complexity of the cross-platform verification task, which involves two players Alice and Bob who have k copies of unknown states ρ,σ, is settled, showing that the success achieved by shadow tomography, for sample-efficiently learning the properties of a single system, cannot be generalized to the distributed setting.
Classical shadows with Pauli-invariant unitary ensembles
The class of Pauli-invariant unitary ensembles that are invariant under multiplication by a Pauli operator are considered, which pave the way for more efficient or robust protocols for predicting important properties of quantum states.
Avoiding Barren Plateaus Using Classical Shadows
This work proposes a general algorithm to avoid barren plateaus in the initialization and throughout the optimization, and demonstrates that decreasing the gradient step size, guided by the entropies allows to avoid WBPs during the optimization process.
Toward Instance-Optimal State Certification With Incoherent Measurements
The first bounds of this nature for the quantum setting are given, showing (up to log factors) that the copy complexity for state certification using nonadaptive incoherent measurements is essentially given by the copy simplicity for mixedness testing times the fidelity between σ and the maximally mixed state.
Revisiting dequantization and quantum advantage in learning tasks
This research presents a probabilistic simulation of the black hole and its environment that allows for the simulation of supermassive black holes and their environment to be studied in detail.
Improving the efficiency of learning-based error mitigation
An error mitigation scheme that builds on Clifford data regression (CDR) is introduced that improves the frugality by carefully choosing the training data and exploiting the symmetries of the problem.


Entanglement is Necessary for Optimal Quantum Property Testing
It is shown that with independent measurements, $\Omega(d^{4/3}/\epsilon^{2})$ is necessary, even if the measurements are chosen adaptively, which resolves a question posed in [7].
Information-theoretic bounds on quantum advantage in machine learning
It is proven that for any input distribution D(x), a classical ML model can provide accurate predictions on average by accessing E a number of times comparable to the optimal quantum ML model, and it is proved that the exponential quantum advantage is possible.
Shadow tomography of quantum states
  • S. Aaronson
  • Computer Science
    Electron. Colloquium Comput. Complex.
  • 2017
Surprisingly, this work gives a procedure that solves the problem of shadow tomography by measuring only O( ε−5·log4 M·logD) copies, which means, for example, that the authors can learn the behavior of an arbitrary n-qubit state, on *all* accepting/rejecting circuits of some fixed polynomial size, by measuringonly nO( 1) copies of the state.
Models of Quantum Complexity Growth
The concept of quantum complexity has far-reaching implications spanning theoretical computer science, quantum many-body physics, and high energy physics. The quantum complexity of a unitary
Efficient learning of quantum noise
The results pave the way for noise metrology in next-generation quantum devices, calibration in the presence of crosstalk, bespoke quantum error-correcting codes 10 and customized fault-tolerance protocols 11 that can greatly reduce the overhead in a quantum computation.
Proposal Draft How to learn a quantum state
  • Computer Science
  • 2015
This thesis proposal is focused on designing and analyzing efficient algorithms for natural learning problems dealing with quantum systems represented as mixed states, and focuses mainly on algorithms derived from Schur-Weyl duality.
Provably efficient machine learning for quantum many-body problems
It is proved that classical ML algorithms can efficiently predict ground state properties of gapped Hamiltonian in finite spatial dimensions, after learning from data obtained by measuring other Hamiltonians in the same quantum phase of matter.
Improved Quantum data analysis
A quantum ”Threshold Search” algorithm that requires only O((log2 m)/є2) samples of a d-dimensional state ρ and a Shadow Tomography algorithm, which simultaneously achieves the best known dependence on each parameter m, d, є, which yields the same sample complexity for quantum Hypothesis Selection.
Fast Learning Requires Good Memory: A Time-Space Lower Bound for Parity Learning
  • R. Raz
  • Computer Science, Mathematics
    2016 IEEE 57th Annual Symposium on Foundations of Computer Science (FOCS)
  • 2016
It is proved that any algorithm for learning parities requires either a memory of quadratic size or an exponential number of samples, and an encryption scheme that requires a private key of length n, as well as time complexity of n per encryption/decryption of each bit is provenly and unconditionally secure.
Experimental quantum Hamiltonian learning
With the help of a quantum simulator and Bayesian inference it is possible to determine the unknown Hamiltonian of a quantum system. An experiment demonstrates this using a photonic quantum simulator