Quantum computing

and quantum machine

learning:
Introduction
—
PD Dr. habil. Jeanette Miriam Lorenz
Fraunhofer Institute for Cognitive Systems IKS
&
LMU Munich
09.08.2023

[https://www.umb.ch/blog/news/detail/quantencomputing-fuer-den-praktischen-einsatz-rueckt-naeher]
Seite 2 09.08.2023 © Fraunhofer IKS
Application fields

[https://www.bcg.com/de-
de/publications/2019/quantum-computers-
create-value-when]

Seite 3 09.08.2023
Emerging technology:
Quantum computing

Quantum computing (QC) has the potential to lead to

disruptive changes in many industrial areas:

▪ Simulation of quantum mechanical systems

▪ (Development of new drugs, chemical sector with battery
development,…)

▪ Optimization problems (Logistics, production, pharma,…)

▪ Quantum machine learning (Computer vision, mobility,…)

Seite 4 09.08.2023 © Fraunhofer IKS

Applications in particle physics

Reconstruction of tracks
- Optimization problem or
quantum machine learning

Event selection
- solved by quantum
Lattice QCD machine learning
– fundamental interactions in discretized space

Seite 5 09.08.2023 © Fraunhofer IKS

Is it all just a hype?

„We help companies make better decisions in less time with What quantum computing might be able to do and
quantum-hybrid computing“ what not:

„Quantum computing has the potential to drive the major ▪ QC will not replace classical computers.
breakthroughs needed to help solve the climate crisis.“ ▪ QC are expected to lead to exponential or polynominal speed-up for
certain calculations (or more precisely subroutines).
„Much like artificial intelligence in its early days, the
▪ Academic quantum advantage has been claimed.
reputation of quantum computing has been tarnished by
▪ A practical quantum advantage has not been shown yet.
grand promises and few concrete results. Talk of quantum
computers is often closely flanked by promises of polynomial- ▪ A claim on quantum utility has been made recently.
time solutions to NP-Hard problems and other such
implausible appeals to blind optimism..”

Seite 6 09.08.2023
Plan for the lectures

1. Intro to quantum computing (Jeanette)

2. Fault-tolerant quantum computing (Federico)

3. Introduction to NISQ quantum computing

(Jeanette)
4. Quantum computing for optimization problems
(Federico)
5. Quantum machine learning (Jeanette)

Lectures interleaved with tutorials

(typically first lecture, then tutorial)

Seite 7 09.08.2023 © Fraunhofer IKS

Jeanette
Searches for SUSY + DM, additional Higgs bosons @ ATLAS -> Quantum computing @ Fraunhofer/LMU

2007 2010–2013 2014–2015 2020 Since 2023

Double degree Research stay @ CERN PostDoc @ Excellence Habilitation in Head of Department
(prediploma) physics and 2010 – 2013. Cluster Universe experimental physics „Quantum-enhanced
mathematics @ FAU (Garching) 2020 – since then AI“ @ Fraunhofer IKS
Erlangen associate professor @
LMU Munich.

Senior scientist @
PhD @ LMU Munich 2014 Junior research group
Fraunhofer IKS 2021 –
Diploma in physics LMU in particle physics (ATLAS leader/assistant
2022 Switched association @
Munich (2010) – diploma Collaboration, searches for professor @ LMU Munich
& project lead quantum LMU Munich to theoretical
thesis in particle physics. SUSY particles). (SUSY searches, DM
computing. physics
searches, additional Higgs
2010 2014 bosons)
2021 - 2022 2023
2015–2021

Seite 8 09.08.2023 © Fraunhofer IKS

Introduction to quantum computing

1. What is quantum computing?

2. Basics:
a) History
b) Current research questions in quantum
computing
c) Definitions
d) What is a qubit?
3. One and multi-qubit gates, simple algorithms
4. Computational complexity

Seite 9 09.08.2023 © Fraunhofer IKS

because nature isn’t classical, dammit, and if you want to
make a simulation of nature, you’d better make it
quantum mechanical, and by golly it's a wonderful
problem, because it doesn't look so easy.
Richard Feynman, Simulating Physics with Computers, International Journal of Theoretical Physics 21, 467 (1982)

Seite 10 09.08.2023 © Fraunhofer IKS

History of QC

[M. Schuld & F. Petruccione, Machine Learning with Quantum Computers, Springer 2021]

Seite 11 09.08.2023
Research questions

▪ What is quantum information?

▪ Can we build a computer based on quantum systems?

▪ How can we then formulate algorithms on such systems?

▪ What does quantum theory mean for the limits of what is computable?

▪ What distinguishes quantum computers from classical ones?

▪ With first small and noisy quantum computers starting to be available: for what can we use
them? (Now and in perspective)
Research topics of
▪ How can we control their imperfection? Jeanette‘s department

▪ How to organise the interplay between classical and quantum computers?

Seite 12 09.08.2023
A few basic definitions

Qubit: A quantum system associated with two measurable states

Quantum computer (QC): Physical implementation of n qubits with precise control on the evolution of the system

Quantum algorithm (QAlg): Controlled manipulations of a quantum system with subsequent measurement to retrieve
information from the system

Quantum gates: Manipulation that act on one or two, … qubits

A quantum algorithm can be formulated as quantum circuits of elementary gates

Seite 13 09.08.2023
In a nutshell: What is quantum computing?

A classical bit can be either 0 or 1.

0
A Quantum Bit (^= Qubit) is the superposition of two
states |0> und |1> at the same time:

Wikimedia
a |0> + b |1>
with |𝑎|2 + |𝑏|2 = 1

But: When measuring the qubit, only the classical states 0

1 and 1 can be measured, i.e. 0 is measured with a
Qubit: Bloch sphere
probability of |𝑎|2 and 1 with a probability of |𝑏|2 .
Classical bit
A quantum computer returns probabilistic results.

Seite 14 09.08.2023 © Fraunhofer IKS

The formal mathematical definition

A qubit is the fundamental unit of quantum information.

At any given time, it is in a superposition state represented by a linear combination of Dirac vectors |0> and |1> in ℂ2 :

|𝜓 > = 𝑎 |0 > +𝑏 |1 > where |𝑎|2 + |𝑏|2 = 1

And 𝑎, 𝑏 ∈ ℂ

The vectors |0> and |1> form an orthonormal basis of a two-dimensional Hilbert space -> Computational basis.

a and b are called probability amplitudes.

|0> and |1> can be represented as the standard basis states of ℂ2 :

1 0
|0 > = ∈ ℂ2 , |1 > = ∈ ℂ2
0 1

Seite 15 09.08.2023
A qubit in polar form

We can rewrite a qubit in polar form:

𝜓 > = 𝑎 0 > +𝑏 1 > = 𝑟1 𝑒 𝜑1𝑖 0 > +𝑟2 𝑒 𝜑2𝑖 |1 >

Furthermore, we can identify two quantum states if they only differ by a multiple of a complex unit, i.e. by a factor 𝑒 𝑖𝜑 for 0 ≤
𝜑 < 2𝜋.

That means the above qubit is effectively the same as:

𝜑2 − 𝜑1 𝑖
𝜓 > = 𝑟1 0 > +𝑟2 𝑒 |1 >

𝑟1 and 𝑟2 are in ℝ and 𝑟1 2 + 𝑟2 2 = 1.

Seite 16 09.08.2023
Mapping to the Bloch sphere

We can find 0 ≤ 𝜃 ≤ 𝜋 so that

𝜃 𝜃
𝑟1 = cos and 𝑟2 = sin
2 2

With this:

𝜃 𝜃 𝜑𝑖
|𝜓 > = cos |0 > + sin 𝑒 |1 >
2 2

Using this, we can use a non-linear projection to get from the three-
dimensional surface of the hypershere in ℂ2 (that one can think of as
ℝ4 ) to a two-dimensional surface of a Bloch sphere in ℝ2 . The main
point why this does work is because we can ignore global phases.

Seite 17 09.08.2023
Basic operations acting on one qubit

Seite 18 09.08.2023 © Fraunhofer IKS

Two qubits

Orthonormal basis formed by two qubits:

|𝜓 >1 = 𝑎1 |0 >1 + 𝑏1 |1 >1 > where |𝑎1 |2 + |𝑏1 |2 = 1

|𝜓 >2 = 𝑎2 |0 >2 + 𝑏2 |1 >2 > where |𝑎2 |2 + |𝑏2 |2 = 1

The orthonormal basis is then:

|0 >1 ⨂|0 >2 , |0 >1 ⨂|1 >2 , |1 >1 ⨂|0 >2 , |1 >1 ⨂|1 >2

Or in short:

|00> , |01> , |10>, |11>

Seite 19 09.08.2023
Example: Representation of numbers

To display numbers from 0 to 15 in bits, we need 4 bits: Four qubits allow to represent all of these 16 states at the
same time:
0000
|0000> + |0001> + |0010> + |0011> + |0100> + |0101> +
0001 |0110> + |0111> + |1000> + |1001> + |1010> + |1011> +
|1100> + |1101> + |1110> + |1111>
0010
→ Parallelisation of calculations possible.
0011

0100 With n qubits 𝟐𝒏 states can be represented at the same

time.
…

1111

Seite 20 09.08.2023 © Fraunhofer IKS

Basic operations acting on multiple qubits

Multi-qubit gates are required to entangle qubits (and

without entanglement some basic benefit of QC not
available).

Definition:

A 2-qubit state in ℂ2 ⊗ ℂ2 is called entangled if and only if it cannot

be written as tensor product of two 1-qubit kets:

|𝜓1 > ⊗ 𝜓2 > = (𝑎1 |0 >1 + 𝑏1 |1 >1 ⊗ (𝑎2 |0 >2 + 𝑏2 |1 >2 )

If a quantum state is not entangled, it is called separable.

Seite 21 09.08.2023 © Fraunhofer IKS

Measurement of a two-qubit system

▪ Usually measurements in the compuational basis

▪ Measurements are represented by projections onto the possible eigenspaces:

𝑃0 = |0 >< 0| or 𝑃1 = 1 >< 1

▪ E.g., 𝑝 0 = 𝑡𝑟 𝑃0 𝜓 >< 𝜓 = < 𝜓 𝑃0 𝜓 > = |𝑎1 |2

▪ The full oberservable corresponding to the computational basis measurement is the Pauli-z observable:
𝜎𝑧 = 0 >< 0 − 1 >< 1

▪ With the eigenvalues:

+1 for |0>
-1 for |1>

Seite 22 09.08.2023
Measurement in practise

▪ How to obtain a expectation value? -> Sample. Run a QAlg s times. s is called the number of shots.

▪ How many shots are required to obtain an estimate < 𝜎𝑧 > with an error 𝜀 ?
-> Bernoulli experiment

▪ In case of large s and a probability of 𝑝 ≈ 0.5 : Wald interval for < 𝜎𝑧 > = 0:
𝑝ො (1 − 𝑝)
ො
𝜀=𝑧 ∙
𝑠
With 𝑝:Ƹ estimator for the probability
z: confidence level

▪ Ο (𝜀 −2 ) samples required for a given 𝜀 and z

▪ Example: 𝜀 = 0.1 and z = 0.99: s = 167

Seite 23 09.08.2023
Measurement in practise

▪ If 𝑝Ƹ → 0 or 𝑝Ƹ → 1:
The Wald approximation is not valid, instead the Wilson score interval
is required:
1/2
𝑧 𝑝Ƹ 1 − 𝑝Ƹ 𝑧2
𝜀= +
𝑧2 𝑠 4𝑠 2
1+
𝑠

=> For 𝜀 = 0.1 only 27 measurements are required for the same
boundaries as before.

[M. Schuld et al, Machine Learning with Quantum Computers, Springer 2021]

Seite 24 09.08.2023
Why is quantum computing promising?

Quantum computing could potentially result in an increased

computing capacity – thus leads to a more efficient solution
Superposition of states of problems.

Entanglement of states – i.e. multiple qubits are → Simplified processing of complicated datasets, solution of
connected/correlated. currently unsolvable problems.

Interference of qubits – i.e. states interfer → However: It is uncertain when the quality of quantum
Enhancement or reduction of states (see Grover!) computers will be sufficient to fully profit from these
advantages.

Seite 25 09.08.2023 © Fraunhofer IKS

The flow of a quantum computation
Simplified: Superposition/Entanglement + Interference -> Solution

[https://qiskit.org/documentation/qc_intro.html]

Seite 26 09.08.2023
What is a quantum computer?

Different criteria to assess quality of quantum computer Di-Vicenzo criteria:

+ if it is a quantum computer:
1. Scalable qubit system with distinct qubits?

▪ Universality 2. Ability to initialize the state of any qubit to a definite

▪ Universal quantum computer? -> Di Vicenzo criteria state in the computational basis

▪ Fidelity (Quality of qubits) 3. The qubits must hold their states

▪ Scalability (Architecture scalable?) 4. Ability to apply unitary operators to qubit states and
to two qubits at once
▪ Qubits: number, architecture-specific limitations like nearest
5. Ability for ‚strong‘ measurements, i.e. the ability that
neighbor connections
the measurement measures the state of the qubit
▪ Logical connectivity (two-qubit gates possible for all qubits?) for the property being measured

▪ Circuit depth (How many operations possible?)

▪ Cloud access (and availability in general)

Seite 27 09.08.2023
Different QC hardware concept (examples)

Neutral atoms Superconducting qubits Trapped ions

▪ Atom ensemble surrounded by laser system ▪ Superconducting Josephson junctions at ▪ Ionized atoms trapped in electric potentials
forming an magneto-optical trap, cryogenic temperatures. to form a line of qubits.
addressable arrays of atoms. ▪ Low connectivity, many SWAP operations ▪ Can be operated at room temperature,
▪ Requires a specific way of programming needed for highly connected circuits, high connectivity, no runtime environment
(pulser), good connectivity, more native to runtime environments in first attempts and running jobs relatively manuel
QUBO formulations? available

Source:
https://www.aqt.eu/
media-press/, Dieter
Kühl

L. Henriet et al, Quantum Computing with

neutral atoms, arXiv:2006.12326v2

Seite 28 09.08.2023 © Fraunhofer IKS

Example for a superconducting chip design

Ibm_cusco

127 qubits

[Visit via the IBM Lab, e.g.,

https://quantum-
computing.ibm.com/services/res
ources ]

Seite 29 09.08.2023
 Classical computation and
complexity

Seite 30 09.08.2023 © Fraunhofer IKS

Classical computers

Components of a classical computer (simplified):

▪ Processor (Central Processing Unit – CPU)

▪ Graphics Processing Unit (GPU) for specialized calculations
▪ Memory for storing information during the calculation
▪ Storage for long-term preservation of data

Difference CPU and GPU:

▪ CPU good at performing different operations, but then maybe slow

▪ GPU good for a set of highly optimized operations – for these very
fast (matrix multiplication)

Quantum computing to be included in these systems as

Quantum Processing Unit

Seite 31 09.08.2023
Storage of information and operations

and its Potential to Transform Silicon-Based Technology. DNA and RNA

[Abels, Seth & Khisamutdinov, Emil. (2015). Nucleic Acid Computing
Classical storage of information in bits

▪ A bit is either 0 or 1
▪ At a low level a computer works with binary numbers
▪ Example: 1101 = 1 ∙ 23 + 1 ∙ 22 + 1 ∙ 21 + 1 ∙ 20

Nanotechnology. 2. 10.1515/rnan-2015-0003.]
Capacity of storage:
▪ Measured in terabytes, gigabytes, …
▪ 1 gigabyte = 1000 megabytes = … = 109 bytes
▪ 1 byte ^= 8 bits

Operations by using transitors

Seite 32 09.08.2023
Seite 33
09.08.2023
Operations

[Abels, Seth & Khisamutdinov, Emil. (2015). Nucleic Acid Computing

and its Potential to Transform Silicon-Based Technology. DNA and RNA
Nanotechnology. 2. 10.1515/rnan-2015-0003.]
How to realize an addition

Task:

0+0=0

1+ 0 = 1

0+1=1

1 + 1 = 0 carry 1

Seite 34 09.08.2023
How to realize an addition

Task: Which logical gates are needed?

0+0=0

1+ 0 = 1

0+1=1

1 + 1 = 0 carry 1

Almost there!

Seite 35 09.08.2023
How to realize an addition

Task: Which logical gates are needed?

0+0=0

1+ 0 = 1

0+1=1

1 + 1 = 0 carry 1

[Source: https://www.101computing.net/binary-
additions-using-logic-gates/ ]

Almost there!

Seite 36 09.08.2023
The interplay between classical and quantum computers

Control system based

on classical technology

Controls QC

QC

Iteration

Classical computer or
HPC system

Seite 37 09.08.2023
A word on complexity and what we need to achieve

In QC we are interested in achieving at least a

polynominal speedup with respect to classical
algorithms.

How to compare algorithms? – The ‚Big – O‘ notation:

Let f(n) and g(n) be functions from positive integers to positive reals.
We say f = O(g) (which means that „f grows no faster than g“) if
there is a constant c > 0 such that 𝑓 𝑛 ≤ 𝑐 ∙ 𝑔 𝑛 .

Example: Sorting

Sort [7, -2, 0, 3] – How do you do it?

1
In general number of swaps ≤ 𝑛2
2
So O(𝑛2 )

Seite 38 09.08.2023 © Fraunhofer IKS

With the advent of near-term quantum devices, this monolithic reliance on
complexity theory is slowly changing. Empirical studies and proof-of-principle
experiments have to deal with the details of an implementation, and a
constant factor in the runtime, for example, if we need n = 20 qubits or cn =
1,000,000∗ 20 qubits (even if the constant c does not grow with the problem
size), suddenly becomes crucial. This is an exciting development: classical
computer science would be widely decimated (and machine learning hardly
existent) if the only algorithms people are interested in were those for which
we can prove efficient runtimes on paper.
Maria Schuld et al.

Seite 39 09.08.2023 © Fraunhofer IKS

How to define quantum speedup?

Terminology developed by T.F. Rnnow et al. to benchmark QCs and quantum algorithms:

Provable quantum speedup:

▪ Proof required that there cannot be a classical algorithm that performs as well or better than the considered quantum algorithm
▪ This is the case for Grover‘s algorithm -> scales quadratically better than classical, given an oracle to mark the desired state

Strong quantum speedup:

▪ Comparison of the quantum algorithm with the best known classical algorithm
▪ For example Shor‘s algorithm

Common quantum speedup:

▪ Strong quantum speedup relaxed to comparing to best available classical algorithm

Potential quantum speedup:

▪ Only comparing two specific algorithms and just referring to this comparison.

Limited quantum speedup:

▪ Comparison of two conceptually equivalent algorithms.
▪ Example: quantum and classical annealing

Seite 40 09.08.2023 © Fraunhofer IKS

Outlook: A map through fault-tolerant QC

Quantum Fourier Transform Amplitude amplification

Phase estimation

Shor HHL Quantum Counting Grover

Solving systems of Insights in NP- Searching unsorted

Factoring
linear equations hard problems? databases

Seite 41 09.08.2023 © Fraunhofer IKS

Tutorial: the addition on a quantum computer

Tutorials attached to the agenda

Setup:

▪ Either linux installation, virtual box with linux installation, or google colab
▪ Install via pip:
▪ pip install qiskit
▪ pip install pennylane
▪ (and probably a few more things will be requested during the installation process)
▪ Installation commands in google colab have to be preceeded by a !

Seite 42 09.08.2023