Introduction to Quantum Error Correction

Coding and Quantum Machine Learning

Assoc. Prof. Dr. Eng. Khoirul Anwar, S.T., M.Eng.1,2,3,4

1Director, The University Center of Excellence for Advanced Intelligent Communications

(AICOMS), Telkom University, Bandung, Indonesia
2Vice-Chairman of Asia-Paci c Wireless Group (AWG), Bangkok, Thailand, 2019—2025
3Chairman of Technical and Education Activities (TEA), Senior Member

IEEE Indonesia Section, 2020, 2021, 2022

4WP Leader of PKR Kuantum 2.0, Indonesia

E-mail: [email protected]

Presented at The 2022 AICOMS-Q Workshop on Quantum Theory, Development, and Business
Bandung (Hybrid), 8 December 2022
fi
PUI-PT AICOMS Transformation
• AICOMS (formerly AdWiTech) established in
September 2016 and received funding from
national and international funding bodies.

2016 2017 2018 2019 2020 2021 2022

Funding: RP700M Funding: RP2.5B Funding: RP2.5B Funding: RP4B Funding: RP4B Funding: RP4B+
• ASEAN IVO • ASEAN IVO •
Established • EPSRC UK • LPDP Rispro • LPDP Rispro • LPDP Rispro
ASEAN IVO
• Ristekdikti • Ristekdikti • LPDP Rispro • LPDP Rispro
since Sept. • Ristekdikti • Ristekdikti • Kominfo
• Kominfo
2016 • Internal • Insinas • Insinas • PUI-PT
•
•
PUI-PT
Kemendikbud • PUI-PT
• Kemendikbud • BRIN • BRIN 2
AICOMS Current 5 Top Products/Projects
(1) MCRBS (2) High speed train (3) Over-The-Horizon Comm. /Rocket
& 5G-MERDEKA (5G-FRMCS)
Mobile Cognitive
Radio Base Station

image: © K. Anwar, Telkom University, 2020.

(4) Next Generation IoT (5) AI, Quantum Coding & Security image:
nature

image: © K. Anwar, IEEE APCC2021. 3

National and International Recognitions of AICOMS by April 2022

Cluster 1
AICOMS, Telkom University

Others
Cluster 2

Others
Cluster 3

• In 2022, AICOMS is becoming Cluster 1

achieving high quality research targets and
management of laboratory.

• Our research on 5G-FRMCS is recognized

by the international community of FRMCS. source: everybodywiki 4
 Laws of Physical Worlds

1879-1955 1643-1727 1901-1976

• All matter and object motions in this physical world are strictly governed by his three laws of dynamics
known as Newton’s laws of motion.
2
• Groundbreaking mass–energy equation E = mc has been considered as “the world’s most famous
equation” after Newton’s second law of motion F = ma.

• Quantum mechanics (QM) and quantum eld theory describe nature at the smallest scales of energy
levels of atoms and subatomic particles.
5
image: Raymond Lee, Quantum nance, 2020.
fi
fi
Motivation: Succesful Digital Transformations

2021 2022

AICOMS is supporting that unexploited 2018-2022

business opportunities should be lled by
innovative university and startup 2G,
companies that exploit digital technologies. 3G,
4G,
5G,
Support → 6G
IoT,

image: Volker Lang, Digital Fluency, Apress, 2021 6

fi
The Philosophy of Telecommunications
Error correction Energy Business
Bandwidth
EM or quantum Channels

( N0 )
E
2 b 2
C ≈ K ⋅ n ⋅ B log2 1 + m ⋅ R ⋅ | h | bit/s/Hz/m /Joule/Rp

OAM Modulation shaping loss

Massive MIMO

it
lim
> 1Tbps

n
no
an
Sh
• Understanding philosophy for being
better in research and teaching
7
Motivation (1/4): Trend of Telecommunications

• The trend of future

telecommunications will be
on photonics and quantum
radio.
• Certain problems are hard
for Turing machine, but
easy for Quantum machine
• Google claimed in 2019 that
Quantum machine
Sycamore did the task in
200 seconds which would
have taken 10,000 years for
a supercomputer to perform.

3
Information Theory: Philosophy

Image: G. Smith, The theory of quantum information: channels, capacities, and all that, Feb. 2021 9
Information Theory: Channel Capacity

• The capacity is the amount

of stu you cand send per
use of the channel.

• For a truck: tons/load

• For classical channel: bits/

channel use
Image: G. Smith, The theory of quantum information: channels, capacities, and all that, Feb. 2021 10
ff
Information Theory: Quantum Channel Capacity

• Additivity tells us that the resource only interract

trivially.

• The capacity is the amount of stu you cand

send per use of the channel.

• If you send qubits, it is quantum channel capacity

• if you send bits, it is classical channel capacity

Image: G. Smith, The theory of quantum information: channels, capacities, and all that, Feb. 2021 11
ff
Current Interesting Issue: MAC + Entanglement

Image: G. Smith, The theory of quantum information: channels, capacities, and all that, Feb. 2021 12
Maximizing 5G to Prepare 6G

2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030

Expected Released 6G Standard

10
13
Potential Technology Enablers IMT-2030
• 1. AI Communications
• AI-native new air interface
• AI-native radio network
• Radio Network for AI
• AI-driven RAN architecture and optimization
• Explainable AI for RAN
• 2. Integrated Sensing and Communications: sensing-based ultra-high accuracy positioning and localization.
• 3. Convergence of Communications and Computing Architecture: Split computing, Pervasive computing, ubiquitois
computing and data services.

• 4. Integrated Access and Superlink Communications: Sidelink, Radio in THz

• 5. E cient Spectrum Utilization: Spectrum sharing, Broader frequency domain, interference cancellation technique
• 6. Enhance Energy E ciency and Low Power Consumption: Zero-energy IoT, Energy e cient RAN
• 7. Enhanced Services and Applications: self-synthesizing networks, native support real-time communications
• 8. Enhanced Trustworthiness: RAN privacy, Quantum Technology for RAN. 14
ffi
ffi
ffi
 Quantum Communications in 6G

15
Image: Zhang et. al, “6G WIRELESS NETWORKS Vision, Requirements, Architecture, and KeyTechnologies”, IEEE VT Magazine, July 2019.
Quantum in ITU Study and Plan for IMT-2030 (6G)
• The current IMT-2020 (5G)
standard does not address the
issue of quantum computing
but relies on traditional
cryptography.
• While large-scale quantum
computing can be expected to
take longer, it is time to
prepare for the shift to
cryptography that is secure in
the post-quantum world.
• Security algorithms can use
machine learning to identify
attacks and respond to them.
• Physical layer security
techniques can represent
efficient solutions for securing
the most critical networks.
16
Trend of Quantum Machine Learning
• The term “quantum machine learning”
came into use only around 2013 by
Lloyd, Mohseni and Rebentrost.

• Today, quantum machine learning has

established itself as an active sub-
discipline of quantum computing
research, and comes with a range of
sub-areas.

• It distinguishes four approaches of how

to combine quantum computing and
machine learning, depending on whether
one assumes the data to be generated
by a quantum (Q) or classical (C) system,
and if the information processing device
is quantum (Q) or classical (C).

17
When Quantum Meet AI: QML with QC

Measurement Postulate: Pm = ⟨ψ | M*m Mm | ψ⟩

Mm | ψ⟩
Post Measurement State: | ψ′⟩ = image: Walmsley, IC London
Pm

18
Challenge 3: Future Error Correction Coding

Fixed Design Flexible Design

Traditional Density Evolution, Rateless codes,

Design EXIT Chart Rate-compatible codes

Design by Supervised Reinforcement

Learning Learning Learning

19
Pervasive Intelligence and Non-linear Problems
Neural networks are ‘trained’using input/output
pairs to learn the input/output mapping.
xw = y
x1 ⋅ w1 + x2 ⋅ w2 + x3 ⋅ w3 = y
Solution if
everything is
linear
−1
w=x y

https://soujiroseta.tumblr.com/

In fact, not everything is linear. 20

A Vast Amount of Quantum Information
A Vast Amount
A number of than
larger Quantum Information
number atoms in the universe.
Imagine 300 quantum bits! Current Quantum Computer by August 2020

stores all numbers from 1 to

203703597633448608626844568
8409378161051468393665936250
6361404493543812997633367061
83397376 = 2. 1090
A number larger than the number
of atoms in the universe.

image: Volker Lang, Digital Fluency, Apress, 2021

21
 Processors Are Approaching Quantum
Circuit components approach quantum size
Feature size (mm)

The 2-qubit can be in 4 states at the SAME time

α | 00⟩ + β | 01⟩ + γ | 10⟩ + δ | 11⟩

We need 4 parameters to descrive the state

• Moore’s Law* sets limits to classical computation

* ”The number of transistors incorporated in a chip will approximately double every 24 months”, Gordon Moore, Intel Co-founder (1965)
22
Quantum Mechanic and
Quantum Mechanics in the 20 Quantum
Century Information
th

W. K. Heisenberg P. A. M. Dirac, P. Jordan H. Weil J. von R. Feynman A. Holevo

Neumann
Google
AI
Matrix mechanics Theory of Group Quantum
(1925) and transformations. theory. Quantum information
Indetermination measurem theory
principle (1927). ent. (1970s).

… 1900 2000
Time
M. Planck D. Deutsch
A. Einstein, B. Podolski, N. Rosen J. S. Bell
E. Fermi
IBM Q
Black-body
radiation and EPR Paradox (1935).
discrete nature of L. De Broglie, E. Schrödinger Preliminary
atomic and sub- Bell‘s formulation
atomic physics. Theorem of quantum
(1965) computation
Wave mechanics (1924-1926). (1985)
image: Bassoli, “Entanglement-assisted (Quantum) Communication
Entanglement-assisted Networks”,
(Quantum) Communication Networks
IEEE International Conference on Communications 2021
IEEE International (ICC21),
Conference 14-23 June
on Communications 2021
2021 (ICC21) Page 4 23
Virtual // 14-23 June 2021
The History of 3 Important Technology for 6G
Neural Computer
1948 1950 1960 1969 1980 1990 2020
Cybernetics Multilayer NN Back prop, GD SVM Many
& Learning in The Brain Perceptron & XOR Problem Application
s

1900 1959 1984 1994 1995 1996 2017 2019

M. Planck R. Feynman QKD Protocol Shor’s Quantum Grover Quantum Google
BB84 Algorithm Error Algorithm Satellite Sycamore
Correction

1992 2020 2022

OAM can be carried by optical vortex beam Japan, 200 China, 1 Tbps
Gbps
24
Quantum Computing and Machine Learning

• Two in uential quantum algorithms were Shor’s factorisation algorithm and

Grover’s search algorithm.

• From roughly 2013 onwards, quantum machine learning became its own sub-
discipline of quantum computing 25
fl
Quantum Information Theory and Communications
• Venn logic diagram
representing the main
research sub-areas in the
research eld of quantum
communication networks

• Quantum information theory

and error correction are two
important subjects supporting
quantum computing and
communications.

• Quantum mechanics are need

as the fundamental concept.

image: Bassoli, et al., “Quantum Communication Networks, 1st, Springer, Forthcoming, ISBN: 978-3-030- 62938-0. 26
fi
What is Special from Quantum Information (1/2)
Classic: Quantum:
0 or 1 | 0⟩ and | 1⟩

Sequence of 11 bits: Sequence of 11 quantum bits:

(α1 | 0⟩ + β1 | 1⟩)(α2 | 0⟩ + β2 | 1⟩)(α3 | 0⟩ + β3 | 1⟩)(α4 | 0⟩ + β4 | 1⟩)
11111100110 (α5 | 0⟩ + β5 | 1⟩)(α6 | 0⟩ + β6 | 1⟩)(α7 | 0⟩ + β7 | 1⟩)(α8 | 0⟩ + β8 | 1⟩)
(α9 | 0⟩ + β9 | 1⟩)(α10 | 0⟩ + β10 | 1⟩)(α11 | 0⟩ + β11 | 1⟩)

stores only one number: 2022 stores all numbers from 0 to 2047

faster and e cient 27

ffi
Wave and Bit: Classical and Quantum
• Quantum: quantum bit
(qubit) is the
wavefunction of a two-
state quantum system,
which can get in nite
values in from
superposition of values 0
and 1.

• Classical: the bit

represents the
fundamental element,
which only gets two
values {0,1}.
image: Bassoli, “Entanglement-assisted (Quantum) Communication Networks”,
IEEE International Conference on Communications 2021 (ICC21), 14-23 June 2021 28
fi
“Q-Day”
IBM:
50 qubit trapped
ion machine,
2017

Google: 9 qubit trapped ion

device, 2016 Intel: 17 superconducting Google:
qubit chip, 2017 72 qubit
‘Bristlecone’,
2018

?
Sep 21st 2019 China: $11bn quantum
March-June 2020 computing investment

29
image: British Telecom
4
Quantum Computer in Japan 2022 and World

• Japan's Fujitsu and Riken research institute plan to jointly o er

quantum computers to companies in the scal year starting April
2023.

• The computers are expected to be used in nancial forecasting and

the development of new materials and medicines.

• IBM installed its own quantum computer in Kawasaki, in 2021.

Japanese companies have lagged foreign ones in the development of
such computers.

• Google is aiming to commercialize its machine in 2029, with

applications in drug discovery and the development of new
batteries.

• Fujitsu’s computer will have 64 qubits, more than the 53 qubits in

Google's 2019 machine and second only to IBM's 127 qubits
developed in 2021. Fujitsu says it hopes to produce a machine that has
more than 1,000 qubits after April 2026.

• Boston Consulting Group estimates that quantum computing can

create economic bene ts worth $850 billion annually by 2040, due
image: Prime Minister’s O ce of Japan, June 2022, kantei.go.jp. to the development of new materials and other breakthroughs. 30
fi
ffi
fi
fi
ff
IBM Quantum: Development Roadmap Towards 2026

31
The Proposed Quantum Error Correction Codes

32
image: K. Anwar, invited paper in IEEE APCC 2021.
Quantum Gates: Do not Exist in Classical

• Gates for multiple qubits may not be

constructed from a single qubit.

• Measurement is a must in quantum.

image: Bassoli, “Entanglement-assisted (Quantum) Communication Networks”, IEEE International Conference on 33
Communications 2021 (ICC21), 14-23 June 2021
Photon for Quantum Communications
• In the standard model of particle physics, the photon is an elementary particle of
mass zero, charge zero, spin 1 and energy equal to hν, where h is Plancks'
constant and ν the frequency of the classical wave that can be built up by the
quantum mechanical superposition of the same energy photons.

• Electron is both wave and particle.

34
Dual Packet and Wave

35
Simulation of Quantum Photon
• % (c) Dr. Eng. Khoirul Anwar
250
• % Center for Advanced Intelligent
Communications (AICOMS), Telkom 200
University
150
• % Bandung, 13 June 2022
100
•
50
• clear all;
0
• x=-20:0.1:20;
-50
• b=0;
-100
• for ii=1:300
-150
• a=sin((6+ii/500).*x);
-200
• b=a+b;
-250
• end © K. Anwar, 2022
0 50 100 150 200 250 300 350 400
• plot(b)
36
The Meaning of Superposition

α | 4⟩ + β | 5⟩

image: uoregon.edu image: modi ed from wikimedia 37

fi
Schrodinger’s Cat and The Cat State
atom decays or not
q0 = ↵|0i + |1i atom decays or not
<latexit sha1_base64="fHcw5pTpixI/eU7HNo2g89fmYis=">AAACDnicbVA9SwNBEN2L3/ErammzGARBCHcS1EYI2lgqGBPIhTC3mSRL9vbO3T0hXPwFNv4VGwtFbK3t/DdukhPU+GDg8d4MM/OCWHBtXPfTyc3Mzs0vLC7ll1dW19YLG5vXOkoUwyqLRKTqAWgUXGLVcCOwHiuEMBBYC/pnI792i0rzSF6ZQYzNELqSdzgDY6VWYfem5Z74IOIeDF1fgewKpPvUD9DA0PsWWoWiW3LHoNPEy0iRZLhoFT78dsSSEKVhArRueG5smikow5nAu7yfaIyB9aGLDUslhKib6fidO7prlTbtRMqWNHSs/pxIIdR6EAa2MwTT03+9kfif10hM57iZchknBiWbLOokgpqIjrKhba6QGTGwBJji9lbKeqCAGZtg3obg/X15mlwflLzDUvmyXKycZnEskm2yQ/aIR45IhZyTC1IljNyTR/JMXpwH58l5dd4mrTknm9kiv+C8fwG9j5tL</latexit>

poison released
q1 = |0i
<latexit sha1_base64="Lae3EfyKIwApObpKSgWv1XRXYEk=">AAAB9XicbVDLSgNBEOyNrxhfUY9eBoPgKexKUC9C0IvHCOYByRpmJ73JkNnZdWZWCTH/4cWDIl79F2/+jZPHQRMLGoqqbrq7gkRwbVz328ksLa+srmXXcxubW9s7+d29mo5TxbDKYhGrRkA1Ci6xargR2EgU0igQWA/6V2O//oBK81jemkGCfkS7koecUWOlu/u2d/HkthSVXYGknS+4RXcCski8GSnADJV2/qvViVkaoTRMUK2bnpsYf0iV4UzgKNdKNSaU9WkXm5ZKGqH2h5OrR+TIKh0SxsqWNGSi/p4Y0kjrQRTYzoianp73xuJ/XjM14bk/5DJJDUo2XRSmgpiYjCMgHa6QGTGwhDLF7a2E9aiizNigcjYEb/7lRVI7KXqnxdJNqVC+nMWRhQM4hGPw4AzKcA0VqAIDBc/wCm/Oo/PivDsf09aMM5vZhz9wPn8AlcOR7g==</latexit>

or not
© K. Anwar, 2021
q2 = |0i cat alive or dead
<latexit sha1_base64="ml3MfFtpMBFib5pl3UHYKJWNfEk=">AAAB9XicbVDLSgNBEOz1GeMr6tHLYBA8hd0Q1IsQ9OIxgnlAsobZSScZMju7zswqIeY/vHhQxKv/4s2/cZLsQRMLGoqqbrq7glhwbVz321laXlldW89sZDe3tnd2c3v7NR0limGVRSJSjYBqFFxi1XAjsBErpGEgsB4MriZ+/QGV5pG8NcMY/ZD2JO9yRo2V7u7bxYsnt6Wo7Akk7VzeLbhTkEXipSQPKSrt3FerE7EkRGmYoFo3PTc2/ogqw5nAcbaVaIwpG9AeNi2VNETtj6ZXj8mxVTqkGylb0pCp+ntiREOth2FgO0Nq+nrem4j/ec3EdM/9EZdxYlCy2aJuIoiJyCQC0uEKmRFDSyhT3N5KWJ8qyowNKmtD8OZfXiS1YsE7LZRuSvnyZRpHBg7hCE7AgzMowzVUoAoMFDzDK7w5j86L8+58zFqXnHTmAP7A+fwBl1KR7w==</latexit>

image: britannica

image: wikipedia
| i = ↵|000i + |111i
<latexit sha1_base64="8UAWrahwh1haXbb2y7fHBtTHRuc=">AAACHXicbZDLSgMxFIYzXmu9VV26CRZBEMqMFHUjFN24rGAv0CnlTHrahmYyQ5IRStsXceOruHGhiAs34tuY3kBbfwj8fOccTs4fxIJr47rfztLyyuraemojvbm1vbOb2dsv6yhRDEssEpGqBqBRcIklw43AaqwQwkBgJejejOqVB1SaR/Le9GKsh9CWvMUZGIsamfzAjzX3Fci2QHrlg4g7QAeu687YqR+gscjzvBlqZLJuzh2LLhpvarJkqmIj8+k3I5aEKA0ToHXNc2NT74MynAkcpv1EYwysC22sWSshRF3vj68b0mNLmrQVKfukoWP6e6IPoda9MLCdIZiOnq+N4H+1WmJal/U+l3FiULLJolYiqInoKCra5AqZET1rgClu/0pZBxQwYwNN2xC8+ZMXTfks553n8nf5bOF6GkeKHJIjckI8ckEK5JYUSYkw8kieySt5c56cF+fd+Zi0LjnTmQPyR87XD215oNw=</latexit>

• In a large steel box, we place a small amount of

2
radioactive material that has a α probability of
having a single atom decay and emitting a particle 0
<latexit sha1_base64="IHx8NzVYcvP7tDfx1AWuopQzW7I=">AAACHnicbZBNS8NAEIY3ftb6FfXoZbGIglASqR8XoejFo4LVQhPKZDttFzebsLsRSvSXePGvePGgiOBJ/43bWkWrLyy8PDPD7LxRKrg2nvfujI1PTE5NF2aKs3PzC4vu0vK5TjLFsMYSkah6BBoFl1gz3AispwohjgReRJdH/frFFSrNE3lmeimGMXQkb3MGxqKmu3MdpJpvBApkRyA9CECkXaDXnud9sa0gQmOR73+jplvyyt5A9K/xh6ZEhjppuq9BK2FZjNIwAVo3fC81YQ7KcCbwphhkGlNgl9DBhrUSYtRhPjjvhq5b0qLtRNknDR3QnxM5xFr34sh2xmC6erTWh//VGplp74c5l2lmULLPRe1MUJPQfla0xRUyI3rWAFPc/pWyLihgxiZatCH4oyf/NefbZX+3XDmtlKqHwzgKZJWskU3ikz1SJcfkhNQII7fknjySJ+fOeXCenZfP1jFnOLNCfsl5+wDVv6EM</latexit>

| i = ↵|000i + |110i
per hour.

• We also add a Geiger counter that can detect that Radioactive: | keep⟩ = | 0⟩, | released⟩ = | 1⟩
single emission, plus a connected hammer that
can smash open a closed vial of cyanide poison. Poison: | keep⟩ = | 0⟩, | released⟩ = | 1⟩
• If the Geiger counter detects anything, the hammer Cat: | alive⟩ = | 0⟩, | dead⟩ = | 1⟩
swings and cyanide is released into the air.
38
A Simple MATLAB Programming for Quantum Encoder
Prove this circuit: ket0=[1;0];
ket00=kron(ket0,ket0);
ket000=kron(ket00,ket0);
H=1/sqrt(2)*[1 1;1 -1];
HI=kron(H,eye(2));
HII=kron(HI,eye(2));
A=HII*ket000;
X= [0 1;1 0];
CNOT=[eye(2) zeros(2,2); zeros(2,2) X]
CI=kron(CNOT,eye(2)):
B=CI*A;
IC=kron(eye(2),CNOT);
D=IC*B;

Answer:

([0] [0] [0])

1 1 1
(I ⊗ CNOT) ⊗ (CNOT ⊗ I) ⊗ (H ⊗ I ⊗ I) * ⊗ ⊗

IC CI HII ket000 39
 Classification of Stabilizer-based Quantum ECC

40
image: L. Hanzo
Revisit The Classical Repetition and SPC Codes 3a
Repetition Codes R = 1/3 a a
G = [ 1 1 1]
a a

[ 1 0 1]
1 1 0 © K. Anwar, 2020
H=
a a

Single Parity Check (SPC) Codes R = 2/3

2a

[0 1 1]
1 0 1 a a
G=
© K. Anwar, 2020 2b

H = [ 1 1 1] b b

• The basic understanding of matrix G and a⊕b a⊕b

H is important to develop quantum error
correction codes (QECC).
41
Schrodinger’s Cat and The Cat State
atom decays or not
q0 = ↵|0i + |1i atom decays or not
<latexit sha1_base64="fHcw5pTpixI/eU7HNo2g89fmYis=">AAACDnicbVA9SwNBEN2L3/ErammzGARBCHcS1EYI2lgqGBPIhTC3mSRL9vbO3T0hXPwFNv4VGwtFbK3t/DdukhPU+GDg8d4MM/OCWHBtXPfTyc3Mzs0vLC7ll1dW19YLG5vXOkoUwyqLRKTqAWgUXGLVcCOwHiuEMBBYC/pnI792i0rzSF6ZQYzNELqSdzgDY6VWYfem5Z74IOIeDF1fgewKpPvUD9DA0PsWWoWiW3LHoNPEy0iRZLhoFT78dsSSEKVhArRueG5smikow5nAu7yfaIyB9aGLDUslhKib6fidO7prlTbtRMqWNHSs/pxIIdR6EAa2MwTT03+9kfif10hM57iZchknBiWbLOokgpqIjrKhba6QGTGwBJji9lbKeqCAGZtg3obg/X15mlwflLzDUvmyXKycZnEskm2yQ/aIR45IhZyTC1IljNyTR/JMXpwH58l5dd4mrTknm9kiv+C8fwG9j5tL</latexit>

poison released
q1 = |0i
<latexit sha1_base64="Lae3EfyKIwApObpKSgWv1XRXYEk=">AAAB9XicbVDLSgNBEOyNrxhfUY9eBoPgKexKUC9C0IvHCOYByRpmJ73JkNnZdWZWCTH/4cWDIl79F2/+jZPHQRMLGoqqbrq7gkRwbVz328ksLa+srmXXcxubW9s7+d29mo5TxbDKYhGrRkA1Ci6xargR2EgU0igQWA/6V2O//oBK81jemkGCfkS7koecUWOlu/u2d/HkthSVXYGknS+4RXcCski8GSnADJV2/qvViVkaoTRMUK2bnpsYf0iV4UzgKNdKNSaU9WkXm5ZKGqH2h5OrR+TIKh0SxsqWNGSi/p4Y0kjrQRTYzoianp73xuJ/XjM14bk/5DJJDUo2XRSmgpiYjCMgHa6QGTGwhDLF7a2E9aiizNigcjYEb/7lRVI7KXqnxdJNqVC+nMWRhQM4hGPw4AzKcA0VqAIDBc/wCm/Oo/PivDsf09aMM5vZhz9wPn8AlcOR7g==</latexit>

or not
© K. Anwar, 2021
q2 = |0i cat alive or dead
<latexit sha1_base64="ml3MfFtpMBFib5pl3UHYKJWNfEk=">AAAB9XicbVDLSgNBEOz1GeMr6tHLYBA8hd0Q1IsQ9OIxgnlAsobZSScZMju7zswqIeY/vHhQxKv/4s2/cZLsQRMLGoqqbrq7glhwbVz321laXlldW89sZDe3tnd2c3v7NR0limGVRSJSjYBqFFxi1XAjsBErpGEgsB4MriZ+/QGV5pG8NcMY/ZD2JO9yRo2V7u7bxYsnt6Wo7Akk7VzeLbhTkEXipSQPKSrt3FerE7EkRGmYoFo3PTc2/ogqw5nAcbaVaIwpG9AeNi2VNETtj6ZXj8mxVTqkGylb0pCp+ntiREOth2FgO0Nq+nrem4j/ec3EdM/9EZdxYlCy2aJuIoiJyCQC0uEKmRFDSyhT3N5KWJ8qyowNKmtD8OZfXiS1YsE7LZRuSvnyZRpHBg7hCE7AgzMowzVUoAoMFDzDK7w5j86L8+58zFqXnHTmAP7A+fwBl1KR7w==</latexit>

image: britannica

image: wikipedia
| i = ↵|000i + |111i
<latexit sha1_base64="8UAWrahwh1haXbb2y7fHBtTHRuc=">AAACHXicbZDLSgMxFIYzXmu9VV26CRZBEMqMFHUjFN24rGAv0CnlTHrahmYyQ5IRStsXceOruHGhiAs34tuY3kBbfwj8fOccTs4fxIJr47rfztLyyuraemojvbm1vbOb2dsv6yhRDEssEpGqBqBRcIklw43AaqwQwkBgJejejOqVB1SaR/Le9GKsh9CWvMUZGIsamfzAjzX3Fci2QHrlg4g7QAeu687YqR+gscjzvBlqZLJuzh2LLhpvarJkqmIj8+k3I5aEKA0ToHXNc2NT74MynAkcpv1EYwysC22sWSshRF3vj68b0mNLmrQVKfukoWP6e6IPoda9MLCdIZiOnq+N4H+1WmJal/U+l3FiULLJolYiqInoKCra5AqZET1rgClu/0pZBxQwYwNN2xC8+ZMXTfks553n8nf5bOF6GkeKHJIjckI8ckEK5JYUSYkw8kieySt5c56cF+fd+Zi0LjnTmQPyR87XD215oNw=</latexit>

• In a large steel box, we place a small amount of

2
radioactive material that has a α probability of
having a single atom decay and emitting a particle 0
<latexit sha1_base64="IHx8NzVYcvP7tDfx1AWuopQzW7I=">AAACHnicbZBNS8NAEIY3ftb6FfXoZbGIglASqR8XoejFo4LVQhPKZDttFzebsLsRSvSXePGvePGgiOBJ/43bWkWrLyy8PDPD7LxRKrg2nvfujI1PTE5NF2aKs3PzC4vu0vK5TjLFsMYSkah6BBoFl1gz3AispwohjgReRJdH/frFFSrNE3lmeimGMXQkb3MGxqKmu3MdpJpvBApkRyA9CECkXaDXnud9sa0gQmOR73+jplvyyt5A9K/xh6ZEhjppuq9BK2FZjNIwAVo3fC81YQ7KcCbwphhkGlNgl9DBhrUSYtRhPjjvhq5b0qLtRNknDR3QnxM5xFr34sh2xmC6erTWh//VGplp74c5l2lmULLPRe1MUJPQfla0xRUyI3rWAFPc/pWyLihgxiZatCH4oyf/NefbZX+3XDmtlKqHwzgKZJWskU3ikz1SJcfkhNQII7fknjySJ+fOeXCenZfP1jFnOLNCfsl5+wDVv6EM</latexit>

| i = ↵|000i + |110i
per hour.

• We also add a Geiger counter that can detect that Radioactive: | keep⟩ = | 0⟩, | released⟩ = | 1⟩
single emission, plus a connected hammer that
can smash open a closed vial of cyanide poison. Poison: | keep⟩ = | 0⟩, | released⟩ = | 1⟩
• If the Geiger counter detects anything, the hammer Cat: | alive⟩ = | 0⟩, | dead⟩ = | 1⟩
swings and cyanide is released into the air.
42
[[3,1,1]] Quantum Repetition Codes Bit Flip 2
[ 0 0 0 | 0 1 1]
g1 = ZZI 0 0 0 | 1 1 0
H=
g2 = IZZ
H = [Hx | Hz]
[0 0]
T T 0 0
HxHz + HzHx = 04 mod 2 =
d

∑ (n) (1 − p)
d/2
| 0⟩L = | 000⟩ | 1⟩L = | 111⟩ Pe =
d n d−n
p (1 − p) ≈
p 2

n=0
| i
<latexit sha1_base64="jm85o2FRK0yybzX3QsIegyv22Wk=">AAAB9HicbVDLSgNBEOyNrxhfUY9eBoPgKexKUI9BLx4jmAdklzA76U2GzM6uM7OBEPMdXjwo4tWP8ebfOHkcNLGgoajqprsrTAXXxnW/ndza+sbmVn67sLO7t39QPDxq6CRTDOssEYlqhVSj4BLrhhuBrVQhjUOBzXBwO/WbQ1SaJ/LBjFIMYtqTPOKMGisFT36qua+o7AkknWLJLbszkFXiLUgJFqh1il9+N2FZjNIwQbVue25qgjFVhjOBk4KfaUwpG9Aeti2VNEYdjGdHT8iZVbokSpQtachM/T0xprHWozi0nTE1fb3sTcX/vHZmoutgzGWaGZRsvijKBDEJmSZAulwhM2JkCWWK21sJ61NFmbE5FWwI3vLLq6RxUfYuy5X7Sql6s4gjDydwCufgwRVU4Q5qUAcGj/AMr/DmDJ0X5935mLfmnMXMMfyB8/kDyZySHg==</latexit>

© K. Anwar, 2022
|0i
<latexit sha1_base64="Rp9p3w4hCxG08xa/QEmcuy8N5gM=">AAAB8XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE1GPRi8cK9gPbUDbbSbt0swm7G6HE/gsvHhTx6r/x5r9x2+agrQ8GHu/NMDMvSATXxnW/ncLK6tr6RnGztLW9s7tX3j9o6jhVDBssFrFqB1Sj4BIbhhuB7UQhjQKBrWB0M/Vbj6g0j+W9GSfoR3QgecgZNVZ6eHK7isqBQNIrV9yqOwNZJl5OKpCj3it/dfsxSyOUhgmqdcdzE+NnVBnOBE5K3VRjQtmIDrBjqaQRaj+bXTwhJ1bpkzBWtqQhM/X3REYjrcdRYDsjaoZ60ZuK/3md1IRXfsZlkhqUbL4oTAUxMZm+T/pcITNibAllittbCRtSRZmxIZVsCN7iy8ukeVb1Lqrnd+eV2nUeRxGO4BhOwYNLqMEt1KEBDCQ8wyu8Odp5cd6dj3lrwclnDuEPnM8fERuQiA==</latexit>

|0i
<latexit sha1_base64="Rp9p3w4hCxG08xa/QEmcuy8N5gM=">AAAB8XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE1GPRi8cK9gPbUDbbSbt0swm7G6HE/gsvHhTx6r/x5r9x2+agrQ8GHu/NMDMvSATXxnW/ncLK6tr6RnGztLW9s7tX3j9o6jhVDBssFrFqB1Sj4BIbhhuB7UQhjQKBrWB0M/Vbj6g0j+W9GSfoR3QgecgZNVZ6eHK7isqBQNIrV9yqOwNZJl5OKpCj3it/dfsxSyOUhgmqdcdzE+NnVBnOBE5K3VRjQtmIDrBjqaQRaj+bXTwhJ1bpkzBWtqQhM/X3REYjrcdRYDsjaoZ60ZuK/3md1IRXfsZlkhqUbL4oTAUxMZm+T/pcITNibAllittbCRtSRZmxIZVsCN7iy8ukeVb1Lqrnd+eV2nUeRxGO4BhOwYNLqMEt1KEBDCQ8wyu8Odp5cd6dj3lrwclnDuEPnM8fERuQiA==</latexit>

|0i
<latexit sha1_base64="Rp9p3w4hCxG08xa/QEmcuy8N5gM=">AAAB8XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE1GPRi8cK9gPbUDbbSbt0swm7G6HE/gsvHhTx6r/x5r9x2+agrQ8GHu/NMDMvSATXxnW/ncLK6tr6RnGztLW9s7tX3j9o6jhVDBssFrFqB1Sj4BIbhhuB7UQhjQKBrWB0M/Vbj6g0j+W9GSfoR3QgecgZNVZ6eHK7isqBQNIrV9yqOwNZJl5OKpCj3it/dfsxSyOUhgmqdcdzE+NnVBnOBE5K3VRjQtmIDrBjqaQRaj+bXTwhJ1bpkzBWtqQhM/X3REYjrcdRYDsjaoZ60ZuK/3md1IRXfsZlkhqUbL4oTAUxMZm+T/pcITNibAllittbCRtSRZmxIZVsCN7iy8ukeVb1Lqrnd+eV2nUeRxGO4BhOwYNLqMEt1KEBDCQ8wyu8Odp5cd6dj3lrwclnDuEPnM8fERuQiA==</latexit>

Z Z
<latexit sha1_base64="SnYQS61M9QihKlqmdyE6xQ3C+Lw=">AAAB6HicbVDLTgJBEOzFF+IL9ehlIjHxRHYNUY9ELx4hkUeEDZkdemFkdnYzM2tCCF/gxYPGePWTvPk3DrAHBSvppFLVne6uIBFcG9f9dnJr6xubW/ntws7u3v5B8fCoqeNUMWywWMSqHVCNgktsGG4EthOFNAoEtoLR7cxvPaHSPJb3ZpygH9GB5CFn1Fip/tArltyyOwdZJV5GSpCh1it+dfsxSyOUhgmqdcdzE+NPqDKcCZwWuqnGhLIRHWDHUkkj1P5kfuiUnFmlT8JY2ZKGzNXfExMaaT2OAtsZUTPUy95M/M/rpCa89idcJqlByRaLwlQQE5PZ16TPFTIjxpZQpri9lbAhVZQZm03BhuAtv7xKmhdl77JcqVdK1ZssjjycwCmcgwdXUIU7qEEDGCA8wyu8OY/Oi/PufCxac042cwx/4Hz+ALuvjOc=</latexit>

<latexit sha1_base64="SnYQS61M9QihKlqmdyE6xQ3C+Lw=">AAAB6HicbVDLTgJBEOzFF+IL9ehlIjHxRHYNUY9ELx4hkUeEDZkdemFkdnYzM2tCCF/gxYPGePWTvPk3DrAHBSvppFLVne6uIBFcG9f9dnJr6xubW/ntws7u3v5B8fCoqeNUMWywWMSqHVCNgktsGG4EthOFNAoEtoLR7cxvPaHSPJb3ZpygH9GB5CFn1Fip/tArltyyOwdZJV5GSpCh1it+dfsxSyOUhgmqdcdzE+NPqDKcCZwWuqnGhLIRHWDHUkkj1P5kfuiUnFmlT8JY2ZKGzNXfExMaaT2OAtsZUTPUy95M/M/rpCa89idcJqlByRaLwlQQE5PZ16TPFTIjxpZQpri9lbAhVZQZm03BhuAtv7xKmhdl77JcqVdK1ZssjjycwCmcgwdXUIU7qEEDGCA8wyu8OY/Oi/PufCxac042cwx/4Hz+ALuvjOc=</latexit>

| t i = ↵|000i + |111i
<latexit sha1_base64="UdACAh3wR/TgAgHSvSazoE9XXdI=">AAACIHicbZBNSwMxEIazflu/qh69BIsgCGVXivUiiF48KthW6JYym07bYDa7JLNCqf4UL/4VLx4U0Zv+GtO6gl8vBF6emWEyb5Qqacn337yJyanpmdm5+cLC4tLySnF1rW6TzAisiUQl5iICi0pqrJEkhRepQYgjhY3o8nhUb1yhsTLR5zRIsRVDT8uuFEAOtYvV6zC1sk2hAd1TyA94CCrtA7/2ff8L7vAwQnIsCIKctYslv+yPxf+aIDclluu0XXwNO4nIYtQkFFjbDPyUWkMwJIXCm0KYWUxBXEIPm85qiNG2huMDb/iWIx3eTYx7mviYfp8YQmztII5cZwzUt79rI/hfrZlRd781lDrNCLX4XNTNFKeEj9LiHWlQkBo4A8JI91cu+mBAkMu04EIIfp/819R3y8FeuXJWKR0e5XHMsQ22ybZZwKrskJ2wU1Zjgt2ye/bInrw778F79l4+Wye8fGad/ZD3/gGAlaHt</latexit>

|0i
<latexit sha1_base64="Rp9p3w4hCxG08xa/QEmcuy8N5gM=">AAAB8XicbVBNS8NAEJ3Ur1q/qh69LBbBU0lE1GPRi8cK9gPbUDbbSbt0swm7G6HE/gsvHhTx6r/x5r9x2+agrQ8GHu/NMDMvSATXxnW/ncLK6tr6RnGztLW9s7tX3j9o6jhVDBssFrFqB1Sj4BIbhhuB7UQhjQKBrWB0M/Vbj6g0j+W9GSfoR3QgecgZNVZ6eHK7isqBQNIrV9yqOwNZJl5OKpCj3it/dfsxSyOUhgmqdcdzE+NnVBnOBE5K3VRjQtmIDrBjqaQRaj+bXTwhJ1bpkzBWtqQhM/X3REYjrcdRYDsjaoZ60ZuK/3md1IRXfsZlkhqUbL4oTAUxMZm+T/pcITNibAllittbCRtSRZmxIZVsCN7iy8ukeVb1Lqrnd+eV2nUeRxGO4BhOwYNLqMEt1KEBDCQ8wyu8Odp5cd6dj3lrwclnDuEPnM8fERuQiA==</latexit>

Z Z
<latexit sha1_base64="SnYQS61M9QihKlqmdyE6xQ3C+Lw=">AAAB6HicbVDLTgJBEOzFF+IL9ehlIjHxRHYNUY9ELx4hkUeEDZkdemFkdnYzM2tCCF/gxYPGePWTvPk3DrAHBSvppFLVne6uIBFcG9f9dnJr6xubW/ntws7u3v5B8fCoqeNUMWywWMSqHVCNgktsGG4EthOFNAoEtoLR7cxvPaHSPJb3ZpygH9GB5CFn1Fip/tArltyyOwdZJV5GSpCh1it+dfsxSyOUhgmqdcdzE+NPqDKcCZwWuqnGhLIRHWDHUkkj1P5kfuiUnFmlT8JY2ZKGzNXfExMaaT2OAtsZUTPUy95M/M/rpCa89idcJqlByRaLwlQQE5PZ16TPFTIjxpZQpri9lbAhVZQZm03BhuAtv7xKmhdl77JcqVdK1ZssjjycwCmcgwdXUIU7qEEDGCA8wyu8OY/Oi/PufCxac042cwx/4Hz+ALuvjOc=</latexit>

<latexit sha1_base64="SnYQS61M9QihKlqmdyE6xQ3C+Lw=">AAAB6HicbVDLTgJBEOzFF+IL9ehlIjHxRHYNUY9ELx4hkUeEDZkdemFkdnYzM2tCCF/gxYPGePWTvPk3DrAHBSvppFLVne6uIBFcG9f9dnJr6xubW/ntws7u3v5B8fCoqeNUMWywWMSqHVCNgktsGG4EthOFNAoEtoLR7cxvPaHSPJb3ZpygH9GB5CFn1Fip/tArltyyOwdZJV5GSpCh1it+dfsxSyOUhgmqdcdzE+NPqDKcCZwWuqnGhLIRHWDHUkkj1P5kfuiUnFmlT8JY2ZKGzNXfExMaaT2OAtsZUTPUy95M/M/rpCa89idcJqlByRaLwlQQE5PZ16TPFTIjxpZQpri9lbAhVZQZm03BhuAtv7xKmhdl77JcqVdK1ZssjjycwCmcgwdXUIU7qEEDGCA8wyu8OY/Oi/PufCxac042cwx/4Hz+ALuvjOc=</latexit>

Transmiter Channel Receiver 43

Philosophy of Stabilizer
• The state | ψ⟩ is said to be stabilized by operator X1X2 if X1X2 | ψ⟩ = | ψ⟩.
• The state | ψ⟩ is said to be stabilized by operator Z1Z2 if Z1Z2 | ψ⟩ = | ψ⟩.
• Many quantum codes (including CSS codes and the Shor code) can be much
more compactly described using stabilizers than in the state vector description.

• For n = 3, we have stabilizer S = {III, Z1Z2, Z2Z3, Z1Z3}.

• Operator Z1Z2 stabilizes states | 000⟩, | 001⟩, | 110⟩, | 111⟩.
• Operator Z2Z3 stabilizes states | 000⟩, | 100⟩, | 011⟩, | 111⟩.
• Logical operator are the common states | 000⟩ and | 111⟩.

44
How to Detect Error (3 Qubits)

45
Image: J. Ro e
ff
Shor Codes (1/2)

We rst encode the qubit

using the phase ip code:

© K. Anwar, 2022

A B C

46
fi
fl
 Encoder Decoder

8
Ka IIIXXXXXX 47
What is The Minimum Qubits to Protect from A
Single Qubit Error
• Hamming bound for t qubits correction capability, the number of n qubits is obtained from

• Singleton bound is to make sure existence of such codes, we should satisfy

• Therefore, with d = 2t + 1, t = 1, k = 1 we have

n − k ≥ 2(2t + 1 − 1)
n ≥ 4t + k
n ≥ 5 qubits
48
The Proposed Quantum Error Correction Codes

49
image: K. Anwar, invited paper in IEEE APCC 2021.
Quantum Machine Learning

50
Machine Learning for XOR Operation (1)
• XOR is an important function for error A L1
correction coding as the parity check. L3
⊕ C
• In 1969, a single neuron is found failed for a
training to perform XOR operation, because B L2
at least three-layer NN is required.
Tabel XOR as training data set
• In the soft-decoding, the operation of XOR is A B C
performed by a box-plus operation ⊞ as 0 0 0
0 1 1

( 2 )
(L1+L2)
1+e 1 0 1
• L3 = L1 ⊞ L2 = log
e L1 + e L 1 1 0

(1 + e 1 2 )
−|L1+L2|
1+e
= sign(L1) ⋅ sign(L2) ⋅ min( | L1 | , | L2 | ) + log −|L −L | ⇐Machine learning
≈ sign(L1) ⋅ sign(L2) ⋅ min( | L1 | , | L2 | ) ⇐Current hardware
51
Machine Learning for XOR Operation (2)
• We consider three-layer NN. A L1
L3
• Since log-likelihood ratio (LLR) taking both
positive and negative value is considered, we
⊕ C
use the action functions: B L2

• f1(z) = tanh(z)
© K. Anwar, 2022

A y
z1 f1 1
• f2(z) = tanh(z)
• f3(z3) = z3 y2 y3
B z2 f2 z3 f3
• with zi being the logit of the i-th neuron.
• y1, y2, y3 are output of each neuron. 1 1
• We include bias to observe the e ects. 52
ff
Machine Learning for XOR Operation (3)
Let the matrix of weight originally of
f1(z) = tanh(z)

[1 1.2 1.4]
2.3 0.9801
1 1.1 1.2
w1 = z1 f1 f2(z) = tanh(z)
0 A w11 = 1

f3(z) = z

w2
1
=

w1
and

=
1.
. 1

1
1
w2 = [1.1 1.2 1.3] = 2.6 0.9890 3.5649
w 12
y3 = 3.5649
1
B w22 = 1.2
z2 f2 w2 = 1.2
z3 f3
We try data 2, i.e.,
A=0, B=1, C=1.

.2
© K. Anwar, 2022 . 3
=1 =
1 e = 1 − y3 = − 2.5649

4
The error is w3
13

1.
w

=
1 − y3 = − 2.5649. 1 1
23
w

53
Machine Learning for XOR Operation (4)
The learning rate is α = 0.1
f1(z) = tanh(z)

[1 1.1933 1.3933] 0
e1 = − 2.8214
ℓ=1 1 1.0889 1.1889 −0.1112
w1 =
A w11 = 1−0
v
δ1 1
f2(z) = tanh(z)
f3(z) = z

w2
1
1

w1
and 1 1

1
. 0

=
0
−

1.
. 1
w2 = [0.8486 0.9463 1.0435]

1
1
= −0.0672 e2 = − 3.0779 δ = − 2.5649
2
w 1
y3 = 3.5649
1
B v w2 = 1.2
δ f3
[ ]
δ2 2
w22 = 1.2−0.0067
ℓ=1 0 −0.0111 −0.0111
Δw1 =
0 −0.0067 −0.0067

111
0.0
.2 −

7
ℓ=1
Δw2 = [−0.2514 −0.2537 −0.2565] © K. Anwar, 2022 . 3

06
=1
1 e = 1 − y3 = − 2.5649

0
=

0.
4−
1 w3
Δw2 = αδy1
13

1.
w

1 1
=
The error is
23

= 0.1 ⋅ −2.5649 ⋅ 0.9801

w

= − 0.2514 1 − y3 = − 2.5649. 54
 Machine Learning for XOR Operation (5)
After the training with 1600 iterations
f1(z) = tanh(z)
[0.7398 0.7450 −0.7286]
2.4937 2.6152 −0.3243 2.2909 0.9797
w1 =
z1 f1 f2(z) = tanh(z)
0 A w11 = 2.4937

and f3(z) = z

w2

w1
1
=
w2 = [1.5922 −1.6581 −0.5328]

=
0.
7

1.
39
2

5
5

92
1

8
0.2

. 6

2
2 0.0164
= 0.0164 1
w 1 2
y3 = 1
B z2 f2 z3 f3
0

1 w22 = 0.7450 w2 = − 1.6581

-0.2

243
2 8

0.3
53
Error

-0.4
© K. Anwar, 2022 0 .
e = 1 − y3 = 0

=−

86
−
=

72
0.
w3
13
-0.6

−
-0.8 1 w
=
23 1
-1 55
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Epoch
Machine Learning for XOR Operation (6)
0.2 0.2

Tanh+linear+bias
Tanh+bias

0 Tanh 0
Sigmoid

-0.2 -0.2

Error
Error

-0.4 -0.4

-0.6 -0.6

Tanh+linear+bias
-0.8 -0.8
Tanh+bias
Tanh
Sigmoid
-1 -1
0 500 1000 1500 2000 2500 3000 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Epoch Epoch #10 4

• NN with tanh+linear+bias is faster in training.

• Sigmoid is not good for Intelligent XOR operation.
56
Our Developed Quantum Machine Learning for 5G and 6G

• A variational circuit can be used as a deterministic

machine learning model that can be used for
supervised learning.

• A popular model: U(x,θ) consists of a data embedding

block S(x) and a parametrised block W(θ).

• An input x is encoded into a quantum state, which

gets processed by a variational circuit.

• The expectation value of a measurement is

interpreted as the output of the model
57
 Input Encoding

• Many quantum algorithms—such as the matrix inversion routine introduced below

—can be understood as strategies of transforming information from one kind of
encoding to the other.

image: Maria Schuld, 2021 58

Quantum Machine Learning for 5G NR CBPSK Demapper
Qubit state Amplitude Step B Amplitude Step C Amplitude Step D 1
q1 q2 q3 q4 x(i) = [(1 − 2b(i)) + j(1 − 2b(i))]
2
0 0 0 0 1/ 2 1.0000 0.7845 Step C is 0
performed using
0 0 0 1 0 0
H gate.
0 0 1 0 1/ 2 0.7500 0.5883
1
0 0 1 1 0 0 (1 + j0.5)
© K. Anwar, 2022 2
0 1 0 0 0 0
0 1 0 1 −1/ 2 0 0
0 1 1 0 0 0
0 1 1 1 −1/ 2 -0.2500 -0.1961 1 q4 is for the label
1 0 0 0 1/ 2 0 0
1 0 0 1 0 0 The real and imaginary parts
1 0 1 0 0.5/ 2 0.2500 0 are assumed as the feature.
p(q4 = 0) = 0.78452 + 0.58832 = 0.9615
1 0 1 1 0 0
← Correct
1 1 0 0 0 0
1 1 0 1 1/ 2 -1.0000 0 p(q4 = 1) = 02 + (−0.1961)2 = 0.0385
1 1 1 0 0 0 We can conclude that the
1 1 1 1 0.5/ 2 -0.7500 0 bit is 0. 59
Our Developed Quantum Machine Learning for 5G and 6G

• QML can succesfully demap 5G CBPSK and 4-QAM by naturally creating the boundary of the areas.
• QML is potential for future M-QAM in 6G, where M > 1024, of which the classical computation is heavy.60
Quantum Deep Learning
• A generic circuit can be seen as a neural
network, where the inputs are the
components ⟨μi |φ(x )⟩ of the data-
encoding state |φ(x )⟩ in the eigenbasis
n
| μ⟩, i = 1,⋯,2 of the measurement
∑
M= μi | μi⟩⟨μi | .
i

• The gates are then i iiii linear layers,

some of which may be xed while others
are trainable.

• The measurement introduces a non-

linear layer with an “absolute square
2
activation” φ(a) = | a | , and the
weights of the nal layer are the
eigenvalues of the measurement
operator.
61
fi
fi