Comments about the Review article in Nature: Quantum machine learning

Following is a discussion about this Review Article in Nature Vol 549 14 September 2017, by Jacob Biamonte, Peter Wittek, Nicola Pancotti, Patrick Rebentrost, Nathan Wiebe & Seth Lloyd

In the last paragraph I explain my own opinion.

Summary - Header

The field of quantum machine learning explores how to devise and implement quantum software that could enable machine learning that is faster than that of classical computers.
The first step is to implement a machine learning problem on a QC and to compare the result of that same problem on a classical computer.
The result should be the same and the performance better.
The second step is to develop quantum software to make implementation on a QC easier.
Recent work has produced quantum algorithms that could act as building blocks of machine learning programs, but the hardware and software challenges are still considerable.
This looks likes Shor's algorithm. This algorithm was developped in 1994 (See page 189) and the hardware to test in at practical scale is not yet available.


In the past decade, the combination of powerful computers and special purpose information processors capable of implementing deep networks with billions of weights, together with their application to very large datasets, has revealed that such deep learning networks are capable of identifying complex and subtle patterns in data.
No problem. I expect that all these applications are properly tested.
Quantum mechanics is well known to produce a typical patterns in data.
This requires an explanation.
Classical machine learning methods such as deep neural networks frequently have the feature that they both recognize statistical patterns in data and produce data that possess the same statistical patterns: they recognize the patterns that they produce.
This seems rather obvious.
If small quantum information processors can produce statistical patterns that are computational difficult for a classical computer to produce, then perhaps they can also recognize patterns that are equally difficult to recognize classically.
Yes, this is a big If.
The realization of this hope depends on whether efficient quantum algorithms can be found for machine learning.
I agree. First you must study an algorithm that performs machine learning on a classical computer.

Classical machine learning

Linear algabra based quantum machine learning

Quantum principal component analysis

Quantum support vector machines and kernel methods

Deep quantum learning

The essential feature of deep quantum learning is that it does not require a large general purpose quantum computer.
To call this an "essential feature" is misleading. General speaking to solve any problem you need initially a large 'general purpose quantum computer' for development. Ofcourse if you have what want you can also implement that same solution on a special purpose QC.
Quantum annealers are well suited for implementing deep quantum learners and are commercially available.
See also Reflection 1 - Definitions
Quantum Boltzmann machines with more general tunable couplings, capable of implementing universal quantum logic, are currently at the design stage.
This is an interesting, but strange, devellopment. Strange in the sense that one should first implement something

Quantum machine learning for quantum data

Dynamnic compilation

Classical co-processing

Designing and controlling quantum systems

Perspective on future work

However, the execution of quantum algorithms requires quantum hardware not yet available
This is a serious issue

Reflection 1 - Definitions

When you read this article it is very important to agree and assume that a quantum computer always uses the concept of superposition, that means qubits which are in two states simultaneous.
A classical computer does not have this capability. Bit is either 0 or 1.
The same that is true for a QC should also be the case for Quantum annealers



Created: 15 February 2018

Back to my home page Index
Back to Nature comments Nature Index