NISQ
NISQ
1 / 18
Timeline of quantum computing
Feynman Grover
full-scale era
NISQ era
Deutsch Shor VQE
Threshold
2 / 18
NISQs and full-scale fault-tolerant quantum computers
3 / 18
Quantum algorithms not covered in the course
4 / 18
Quantum counting, QAE & QMCI
5 / 18
HHL
6 / 18
The quantum singular-value transform
7 / 18
Overview of quantum algorithms
8 / 18
Overview of quantum algorithms
8 / 18
Quantum machine learning
“Quantum machine learning” is a buzz-word heavy slide title, but what
does it actually mean? Crudely, it can be divided into three categories:
Algorithm
quantum classical
classical quantum
Data
9 / 18
Quantum machine learning (cont.)
Quantum machine learning on classical data refers to the use of
quantum algorithms to enhance the performance of conventional
machine learning algorithms. For example, quantum optimisation
(annealing or QAOA), search (Grover) and / or linear system solving
(HHL) may be called as subroutines by some otherwise classical
machine learning algorithm.
Classical machine learning on quantum data refers to the use of
conventional, classical learning techniques to learn something about
some quantum data. Quantum state tomography is a basic example.
Quantum machine learning on quantum data is, it could be
argued, true quantum machine learning, in the sense that we want
to discern some information from a quantum data-set, which may
not be possible if that quantum data were simply measured and
classical learning applied.
In short, the second and third items differ because, in the former the
quantum state is measured and thus collapsed into classical data on
which classical machine learning is applied, whereas in the latter quantum
operations are applied to the quantum data.
10 / 18
Areas quantum computing is expected to impact
Note that this list is by no-means exhaustive, but is simply here to give a
flavour of an optimistic view of what things are to come.
11 / 18
How many qubits do we need?
1 https://www.nature.com/articles/s41586-019-1666-5
2 e.g. https://www.ft.com/content/154a1cf4-ad07-11e8-94bd-cba20d67390c
3 https://doi.org/10.1017/dce.2022.36
4 https://cacm.acm.org/news/237303-how-quantum-computer-could-break-2048-bit-rsa-encryption-in-8-hours/fulltext
12 / 18
From the NISQ era to the full-scale era
Fault-tolerance is the feature that distinguishes the full-scale era from the
NISQ era, and this will require an error correction overhead estimated to
be in the region 20–1000. That is, it will take 20–1000 physical qubits to
make each “clean” logical qubit.
Qubit fidelities and error correcting codes may well improve, bringing this
number down, but the fact remains that a serious scaling-up of the
number of qubits in a quantum computer is needed to build a
fault-tolerant quantum computer.
This in turn has led some in the quantum computing community to talk
about the need for a “quantum transistor” – a highly scalable physical
realisation for a qubit, that changes the game for quantum computing in
the same way that the transistor did for classical computing.
13 / 18
Types of qubit
There are various proposals for physically realising a qubit, of which the
most promising are superconducting qubits and trapped-ion qubits.
14 / 18
How good is a particular quantum computer?
The total number of qubits tends to grab the headlines, but how good a
particular quantum computer is actually depends on three factors:
Qv = 2(min(n,d))
15 / 18
Quantum volume
The depth term, d, is a function of both the fidelity and the depth
overhead incurred when executing a circuit consisting of random
2-qubit interactions.
Therefore the quantum volume is increased for quantum computers
with higher connectivity, as fewer SWAP gates will be needed to
rearrange interacting qubits to be local, and so the depth overhead
will be smaller.
The presence of the min term in the definition indicates whether the
performance of a given quantum computer is limited by a lack of
qubits or poor fidelity / connectivity of the qubits – i.e., a more
nuanced picture than simply quoting the number of qubits.
Quantum volume has been conceptualised so that it gives a
reasonable benchmark of the general performance of near-term
quantum computers.
However some researchers refute that random circuits are
appropriate for this, and instead assert that it part of the role of
quantum software design to execute algorithms in an efficient
manner, given the physical locality constraints of the hardware.
16 / 18
The state of the art in quantum hardware
17 / 18
Quantum-inspired classical algorithms
18 / 18
Quantum-inspired classical algorithms
18 / 18