By Dr Chris Mansell, Senior Scientific Writer at Terra Quantum

Shown below are summaries of a few interesting research papers in quantum computing and communications that we have seen over the past month.


Title: An Improved Classical Singular Value Transformation for Quantum Machine Learning
Organizations: Massachusetts Institute of Technology; University of Washington
The quantum singular value transformation is a quantum algorithm that unifies a host of other quantum algorithms, including Grover’s algorithm, the quantum Fourier transform and Hamiltonian simulation. In this work, the researchers devised a classical algorithm that improves upon the previous best-known algorithm and is only slower than the quantum algorithm by a polynomial overhead. Their “quantum-inspired” algorithm can quickly solve problems like regression, matrix inversion and the task – encountered by service providers like Netflix – of providing personalised recommendations based on information about prior behaviour. They also introduce four theoretical tools that could be of independent interest. 

Title: Towards provably efficient quantum algorithms for large-scale machine-learning models
Organizations: The University of Chicago; Chicago Quantum Exchange; qBraid Co.; SeQure; Argonne National Laboratory; University of California, Berkeley; Massachusetts Institute of Technology; Free University Berlin
Modern neural networks show breathtaking performance, partly because of how large they are. However, their size also makes them very resource intensive, so an important goal is to make them more efficient. One way to do this is to prune (i.e., delete) some of the connections between the neurons. This is done by setting the weight of the connection to zero, which increases the sparseness of the neural network. In this paper, it is argued that a quantum algorithm for solving nonlinear, dissipative ordinary differential equations could enhance the training of sparse classical neural networks. The researchers back up their claims by benchmarking their approach on a network with 103 million parameters. 

Title: Quantum computing reduces systemic risk in financial networks
Organizations: New York University; University of Toronto
The web of financial obligations that exists between banks can lead to a cascade of bank failures, where the insolvency or illiquidity of one bank can cause others to end up in a similar position. Like the phase transitions that occur in physical systems and like the recent case of Silicon Valley Bank, cascades often happen very abruptly. With enough insight, a regulator could minimally rearrange the cross-holdings of the banks so that the transition won’t occur in ordinary circumstances but only in as rare circumstances as possible. However, it is very computationally challenging to figure out how to do this in a realistic and effective way. The authors of this paper compared classical and quantum partitioning algorithms on synthetic and real data. Using D-wave quantum annealers, they found that the quantum approach could make the financial system more resilient to shocks. 

Title: Exploiting Symmetry in Variational Quantum Machine Learning
Organizations: Freie Universität Berlin; Porsche Digital GmbH; Fraunhofer Heinrich Hertz Institute; Helmholtz-Zentrum Berlin für Materialien und Energie
Embedding classical data into present-day quantum computers is a challenging task because it needs to be done in a practical, scalable and useful way. Variational re-uploading methods are a popular approach and this paper investigates how they can be combined with ideas from representation theory, an extremely important area of mathematics that connects abstract algebra (e.g., group theory) with linear algebra. The paper finds that for two machine learning problems, employing a symmetry-aware gate set leads to noteworthy increases in the generalisation performance. Similarly constructed gate sets are also shown to give advantages when applied to variational quantum eigensolvers.

Title: Quantum computation for periodic solids in second quantization
Organizations: Riverlane; Johnson Matthey Technology Centre
This work is about a quantum algorithm for calculating the ground-state energy of electrons within crystalline solids. The researchers took the translational symmetry of these periodic materials into account when they were considering which basis set to use to represent the electron interactions. The best choice for their algorithm was shown to be Wannier functions. Since the algorithm needs to be run on an error-corrected quantum computer, they carefully estimated the number of logic gates it would require for certain tasks. In particular, they looked at two industrially relevant catalysts, nickel oxide and palladium oxide, whose ground-states are poorly described by existing theoretical methods. This is an environmentally conscious goal because improving catalysis reduces the amount of energy that a chemical process needs. 

Title: Quantum Deep Hedging
Organizations: QC Ware; JP Morgan Chase
This paper examines two questions on how the practice of deep hedging which reduces risk for a portfolio utilizing data driven models that consider market frictions and trading constraints might be improved with quantum computing. The researchers first examine whether existing classical deep hedging frameworks could be improved using quantum deep learning. Then, using quantum reinforcement learning, they studied whether a new quantum framework could be defined for deep hedging.


Title: Quantum-classical processing and benchmarking at the pulse-level
Organization: Quantum Machines Inc.
Many of today’s most researched quantum computing techniques, from quantum error correction to variational logic gates, require an exquisite level of control over the qubits. This control is mediated by pulse signals and often has to be interleaved with rapid classical computation. In order to improve how well this can be done, benchmarking is essential. In this paper, benchmarks are proposed for the very lowest level of a quantum computer, the pulse level. This has the advantage that the quality of the qubits and the precision of the controlling operations can be separately investigated and optimised. The work makes use of a comprehensive pulse-level language called quantum universal assembly.

Title: Quantum causality emerging in a delayed-choice quantum Cheshire Cat experiment with neutrons 
Organizations: Atominstitut, TU Wien; Institut Laue Langevin; Hokkaido University
One of the first counterintuitive quantum phenomena that an individual might learn is that of superposition. However, this is only the start. It is possible to send a neutron through an interferometer so that one of its properties (such as the spin) follows one path and the neutron itself follows the other path. It is also possible to choose to insert or remove the final beamsplitter of an interferometer long after the particle has entered the apparatus. In the former case, the neutron is called a quantum Cheshire Cat because it is reminiscent of the cat in ‘Alice in Wonderland’ that can disappear while leaving its smile behind. The latter set-up is called a delayed choice experiment and it is just as weird because the particle seems to start behaving differently depending on a choice that isn’t made until after the behavior has changed. A new experiment has now shown that both of these wonderfully bizarre phenomena can occur together.

Title: Noisy intermediate-scale quantum computers
Organizations: Southern University of Science and Technology; International Quantum Academy; University of Science and Technology of China; RIKEN; University of Michigan
Noisy intermediate-scale quantum computers need no introduction but since there is so much research into them, they do need a review. Written by numerous specialists and referencing over eight hundred papers, this article comprehensively discusses the milestones and breakthroughs as well as the incremental advances and markers of more gradual progress. After a section on quantum algorithms, the paper looks at each of the major hardware platforms in turn, explaining their unique strengths and weaknesses and stating their latest fidelities and qubit counts. The review is especially valuable to the quantum technology community because it is open access.  

Title: Experimental Activation of Strong Local Passive States with Quantum Information
Organizations: University of California, Berkeley; Miller Institute for Basic Research in Science; University of Waterloo; Perimeter Institute for Theoretical Physics
About 15 years ago, Masahiro Hotta devised a scheme to quantum-mechanically teleport energy. Just like the original scheme to teleport the quantum state of one qubit to another qubit, this scheme requires local operations and classical communication. In this peer-reviewed paper, for the first time, the quantum energy teleportation (QET) protocol has been experimentally implemented in a nuclear magnetic resonance system. (A slightly subsequent preprint by Kazuki Ikeda describes a similar QET experiment in an IBM superconducting quantum computer.) In terms of theoretical applications, it is potentially useful for understanding black holes and from a practical perspective, QET could possibly be used alongside procedures that algorithmically cool qubits.

March 31, 2023