mod04lec19 - NISQ-era quantum algorithms

Summary

In this video, Shesha Raghunathan from IBM Systems discusses the evolution and importance of variational quantum algorithms in the NISQ (Noisy Intermediate-Scale Quantum) era. Highlighting the limitations of current quantum hardware, such as noise and qubit count, Raghunathan explains how variational algorithms like VQE and QAOA are structured to fit within these constraints. These hybrid algorithms leverage classical hardware for optimization while performing quantum calculations for tasks like energy evaluation. The talk also touches on the potential applications of these algorithms in quantum machine learning, optimization problems, and more, emphasizing the growing interest and development in this field.

Takeaways

• 📚 The speaker, Shesha Raghunathan, is an IBM Quantum Distinguished Ambassador and leads the ambassador program in India and South Asia.
• 🌟 The talk focuses on modern quantum algorithms, specifically variational algorithms, and their potential applications.
• 🔍 Quantum computing has evolved through three broad generations, starting from Richard Feynman's conceptualization in 1981 to the current era of NISQ (Noisy Intermediate-Scale Quantum) computers.
• ☁️ IBM's release of quantum machines on the cloud in 2016 marked a significant shift, making quantum computing more accessible and sparking increased interest in programming quantum hardware.
• 🔧 The NISQ era, coined by John Preskill in 2007, refers to quantum computers that are noisy and have a limited number of qubits, challenging developers to create algorithms that can provide value despite these constraints.
• 🚀 Variational quantum algorithms, such as VQE and QAOA, emerged in 2014 and gained traction post-2016, aligning with the hardware limitations of the time by focusing on shorter circuit depths.
• 📈 The lifetime of superconducting qubits has exponentially increased over the last 15-20 years, with recent advancements pushing towards millisecond lifetimes, allowing for more complex computations.
• 💡 Variational algorithms are hybrid, utilizing both classical and quantum computing. They are well-suited for the current NISQ hardware, which has constraints on circuit depth and noise levels.
• 🌐 Real-world applications of variational quantum algorithms are being explored, including quantum machine learning, option pricing, and battery optimization.
• 🛠️ The current state of quantum hardware, as of July 2021, shows average qubit lifetimes around 100-120 microseconds, with readout errors dominating over gate errors, emphasizing the need for compact and shallow quantum algorithms.

Q & A

Who is Shesha Raghunathan and what is his role at IBM?

Shesha Raghunathan is part of IBM Systems and works with the Electronic Design Automation team, particularly on timing analysis. He is also an IBM Quantum Distinguished Ambassador, a Qiskit Advocate, and a Technical Ambassador, leading the ambassador program in India and South Asia.

What is the significance of the year 1981 in the context of quantum computing?

The year 1981 is significant because it marks the starting point when Richard Feynman contextualized quantum computing in a more modern form factor.

Why is the year 2016 considered a turning point for quantum computing?

2016 is considered a turning point because that's when IBM put its quantum machine on the cloud for public access, along with a programming platform to program that hardware, which revolutionized the accessibility and programming of quantum computers.

What does the term NISQ stand for and who coined it?

NISQ stands for Noisy Intermediate-Scale Quantum. The term was coined by John Preskill in 2007 to describe quantum computers that are noisy and have a limited number of qubits.

What is the difference between traditional quantum algorithms and those developed for NISQ-era hardware?

Traditional quantum algorithms assume qubits are clean and error-free, whereas NISQ-era algorithms are designed to work with noisy qubits and take into account the hardware's limitations, such as a small number of qubits and noise.

What are variational quantum algorithms and why are they important for NISQ-era hardware?

Variational quantum algorithms are hybrid algorithms that use both quantum and classical computing to solve problems. They are important for NISQ-era hardware because they are designed to be compact and shallow, fitting within the hardware's time and error constraints, and can potentially demonstrate quantum advantage.

What is Quantum Volume and what does it indicate about a quantum computer's capabilities?

Quantum Volume is a measure of the power of a quantum computer, taking into account the number of qubits, the error rates, and the connectivity of the qubits. A higher Quantum Volume indicates a more capable quantum computer.

What is the current state of qubit lifetimes in quantum computers as of the script's reference date?

As of July 10th, 2021, the average qubit lifetime in quantum computers is around 100-120 microseconds, with some experimental qubits reaching milliseconds.

What are the average error rates for the quantum computers mentioned in the script?

The average CNOT error rate is around 0.1 percent, and the readout error rate is around 1 percent for the quantum computers mentioned in the script.

How do variational algorithms fit into the limitations of current quantum hardware?

Variational algorithms are structured as hybrid algorithms with a classical component for optimization and a quantum component for computation. This structure allows them to perform well within the current hardware limitations, such as short qubit lifetimes and error rates.

What are some potential applications of variational quantum algorithms mentioned in the script?

Some potential applications of variational quantum algorithms include quantum machine learning, option pricing, and battery optimization.

Outlines

00:00

🌟 Introduction to Quantum Algorithms and NISQ Era

The speaker, Shesha Raghunathan, introduces the topic of modern quantum algorithms, specifically variational algorithms, and their potential applications. Shesha is an IBM Quantum Distinguished Ambassador and has a background in computer architecture from the University of Southern California. The discussion begins with an overview of quantum computing generations, starting from Richard Feynman's conceptualization in 1981 to the current era, marked by IBM's release of quantum machines on the cloud in 2016. This shift made quantum computing more accessible, leading to increased interest in programming quantum hardware. The era of Noisy Intermediate-Scale Quantum (NISQ) computing is characterized by the presence of noise in qubits and the limited number of qubits available, prompting the need for algorithms that can operate effectively within these constraints.

05:02

📈 Historical Progression and Evolution of Quantum Algorithms

The video script delves into the history of quantum algorithms, starting from Feynman's proposal in 1981 and progressing through various stages of development. It highlights the transition from theoretical exploration to practical applications, such as Shor's algorithm and Grover's algorithm, which demonstrated quantum computing's potential to solve real-world problems. The script also discusses the emergence of variational quantum algorithms like VQE and QAOA in 2014, which gained prominence with the advent of cloud-based quantum computing platforms. These algorithms are designed to work within the limitations of NISQ-era hardware, focusing on shorter circuit depths and hybrid classical-quantum computation. The summary also touches on the exponential increase in qubit lifetimes, reflecting the ongoing improvements in quantum hardware and the potential for more complex quantum computations.

10:03

🛠️ The Relevance of Variational Quantum Algorithms in NISQ

The final paragraph emphasizes the importance of variational quantum algorithms in the context of NISQ-era hardware. It discusses the constraints imposed by the current state of quantum technology, such as the limited lifetime of qubits and the prevalence of noise, which necessitate the development of algorithms that are compact and can be executed within a short time frame. The script introduces the concept of quantum volume, a measure of the power of quantum computers, and provides examples of current IBM quantum machines with their respective specifications. It also highlights the average error rates and the average lifetime of qubits, which are critical factors in determining the feasibility of quantum algorithms. The paragraph concludes by underscoring the significance of variational algorithms in achieving quantum advantage, given the current limitations of quantum hardware, and sets the stage for further exploration of these algorithms and their applications.

Keywords

💡Quantum Algorithms

Quantum algorithms are computational methods that leverage the principles of quantum mechanics to solve problems more efficiently than classical algorithms. In the video, the speaker discusses the evolution of quantum algorithms, particularly focusing on how they've adapted to the limitations of current quantum hardware, such as dealing with noise and qubit limitations. Examples from the script include the shift from traditional quantum algorithms like Shor's algorithm to more modern, variational ones that are better suited for the Noisy Intermediate-Scale Quantum (NISQ) era.

💡Variational Algorithms

Variational algorithms in quantum computing are a class of algorithms that use a hybrid approach, combining quantum and classical computation, to find approximate solutions to optimization problems. They are particularly relevant in the NISQ era due to their ability to work with the noisy and limited qubits available. The video explains how variational algorithms like the Variational Quantum Eigensolver (VQE) and Quantum Approximate Optimization Algorithm (QAOA) have gained traction because they align with the hardware constraints of the current quantum devices.

💡Noisy Intermediate-Scale Quantum (NISQ)

NISQ refers to the current state of quantum computing technology, characterized by devices with a limited number of qubits that are prone to noise and errors. The term was coined by John Preskill in 2007 to describe quantum computers that are not yet powerful enough to be fault-tolerant but are beyond the simple, experimental stage. The video discusses how NISQ devices have spurred the development of variational algorithms that can operate effectively within these constraints.

💡Quantum Volume

Quantum Volume is a metric used to characterize the power of a quantum computer, taking into account not just the number of qubits but also the quality of those qubits, error rates, and connectivity. The video mentions that the quantum volume of the machines discussed is 128, indicating a relatively high capability for these devices. This metric is important because it helps to understand the potential of current quantum hardware for practical applications.

💡Qubit

A qubit, short for quantum bit, is the fundamental unit of quantum information, analogous to the bit in classical computing. Qubits can exist in a superposition of states, allowing quantum computers to process information in ways that classical computers cannot. The video discusses the evolution of qubit technology, noting the increase in qubit lifetimes and the challenges of noise and error rates, which are critical factors in the performance of quantum algorithms.

💡Quantum Advantage

Quantum advantage refers to the point at which quantum computers can solve certain problems more efficiently than classical computers. It is a key goal in the field of quantum computing. The video discusses the pursuit of quantum advantage in the context of NISQ devices, emphasizing the need for algorithms that can solve problems in a compact and efficient manner, exploiting the unique capabilities of quantum hardware.

💡Error Correction

Error correction in quantum computing is the process of detecting and correcting errors that occur during quantum computation. This is crucial because quantum systems are highly susceptible to errors due to their interaction with the environment. The video touches on the history of quantum algorithms and how early work in error correction and fault tolerance laid the groundwork for dealing with the noisy nature of NISQ devices.

💡Hybrid Algorithms

Hybrid algorithms are those that combine quantum and classical computational elements. In the context of quantum computing, this often means running certain parts of an algorithm on a quantum processor and others on a classical computer. The video explains how variational algorithms are hybrid by nature, using classical optimization to guide the quantum computation, which is essential for navigating the limitations of current quantum hardware.

💡Quantum Machine Learning

Quantum machine learning is an emerging field that applies quantum computing techniques to machine learning problems, potentially offering speedups for certain tasks. The video mentions quantum machine learning as one of the potential applications of quantum algorithms, suggesting that the variational approach could be particularly beneficial in this domain.

💡Lifetime

In the context of quantum computing, lifetime refers to the duration for which a qubit can maintain its quantum state before being affected by noise or other errors. The video discusses the exponential increase in qubit lifetimes over the past two decades, which is critical for performing more complex quantum computations. Longer lifetimes allow for more operations to be performed before the qubit state degrades, enhancing the capabilities of quantum algorithms.

Highlights

Introduction to modern quantum algorithms and variational algorithms by Shesha Raghunathan.

Shesha Raghunathan's background in computer architecture and role at IBM Systems.

Three generations of quantum computing: experimental, noisy intermediate-scale quantum (NISQ), and fault-tolerant.

IBM's role in mainstreaming quantum computing by putting quantum machines on cloud in 2016.

The concept of Noisy Intermediate-Scale Quantum (NISQ) era introduced by John Preskill in 2007.

Challenges in programming quantum hardware in the NISQ era due to noise and limited qubit numbers.

Historical development of quantum algorithms from theoretical to practical applications.

Emergence of variational algorithms like VQE and QAOA in 2014, aligning with hardware limitations.

Quantum volume as a measure of quantum computer's power, with IBM machines having a volume of 128.

The importance of qubit lifetime in quantum computing, with current state-of-the-art around 100-120 microseconds.

Error rates in quantum computing, including CNOT and readout errors, and their impact on algorithm design.

Variational quantum algorithms as a hybrid approach fitting current hardware limitations.

Potential applications of variational quantum algorithms in quantum machine learning and optimization.

The necessity for algorithms to solve problems that classical hardware finds hard within the quantum hardware's time budget.

The trend of increasing qubit lifetimes, moving towards milliseconds, and its implications for quantum computing.

The structure of variational algorithms, combining classical optimization with quantum computation.

The future of variational quantum algorithms in the NISQ era and their significance in quantum computing advancements.

Transcripts

[Music]

hi

welcome um in this week we're going to

talk about more modern algorithms

niskira quantum algorithms we're going

to learn about what variational

algorithms are and its potential

applications

my name is shesha raghunathan i'm part

of ibm systems

i work with electronic design automation

team particularly on timing analysis i'm

also an ibm quantum distinguished

ambassador and kisket advocate and a

technical ambassador

i lead the ambassador program in india

south asia

my phd was in computer corner computing

in from university of southern

california in 2010 and i've been with

ibm since 2011. so there are uh

many aspects to the modern quantum

algorithms so in this section what we're

going to learn is an introduction and a

motivation to

why we need these variation of quantum

algorithms how are they different from

the ones that we had before

but before we get started we want to

understand a little bit of the

generations of the broader scope

where we are where we have come from and

where we are heading um as such there

are three broad generations um

that

quantum computing um we can break

quantum computing into three broad uh

generations

um if we take 1981 as the starting point

when

richard feynman

contextualized

quantum computing in the more modern

form factor

uh since

2019 1981 to about 2016

is where we could broadly call it as

experimental why 2016 it's because uh

2016 is when ibm put its quantum machine

on cloud for access

and also a programming platform to

program that hardware

what that did is that people then

started to start accessing these

machines and to start programming it

before then it was mostly in the

experimental mode where it was in some

basement of physics lab or in some

companies

working out of

some

some corner of particular

research labs

but as such the mainstreaming of this

particular quantum computing as a

platform and that too on cloud

revolutionized many things

since then there has been a

transformation in terms of the

interest in programming this hardware

before then programming was less

important and and since 2016 it has

taken a new life of its own

john prescott

in 2007

coined the term noisy intermediate state

quarter

or in short nisk

what that was was that

the question was um currently the

hardware is still evolving the number of

qubits is noise the qubits is noisy and

the number of them is also small

um so he was

looking at qubit systems that are like

50 to 100 in range and they are noisy

the question was in this space can we do

something useful that provides value and

can we demonstrate something of an

advantage

this era is what he called as a noisy

intermediate state quantum

so

what is

the default then

so all the algorithms that was developed

prior to this are many of them

traditional that many of you are aware

of like shores uh bernstein was irony or

any such

doesn't deal with noise per se

so they look at a qubit that is clean

and then they start programming it

even in the 90s a lot of work happened

in terms of

error correction and fault tolerance and

so the concept or the worldview of

programming was that the qubit is clean

we just need to do the algorithm

but now in the niskira qubit is assumed

to be noisy then how do we program it so

that we get maximum out of this hardware

although preschool

contextualized the nisk in the space of

50 to 100 cubits

but we're going to cross that pretty

rapidly pretty soon so

technically it doesn't hold it but now

it has become more a label uh to capture

the space where we are in this noisy

regime and some point later in the

future we're going to transition to what

is eventually a fault tolerant regime

where the qubits can be considered clean

and there is an auto correction or

management of the noise more inherently

done within the hardware and the

software stack and so the programming

part of it the algorithm need not worry

about that but that is little further

out

in the meantime we are banging

right in the niskira and what is it that

we need to do to manage our algorithms

is the question

just a brief history i touched on some

of these things

in the prior chart

so if you look at the history a lot of

investigations in quantum algorithms

started since

feynman's proposal in 1981

initially there was questions regarding

what is the idea

of quantum does it provide any value a

lot of

problems toy problems were created to

demonstrate complexity angle for it does

it provide any additional value over and

above our classical

it went through into the 90s like deuce

george and bernstein was irani are

somewhat of artificial problems

trying to solve some problem but that

brings out a key factor in quantum to

demonstrate its key value proposition

and then we got into some more serious

work like shores and growers that were

solving real problems a lot of

algorithms have been developed since by

no means this is exhaustive but

important thing happened in 2014 a

couple of algorithms came vqe we're

going to learn about that in detail

in this lecture and also qaoa pqa stands

for variational quantum eigensolver qaoa

is a quantum approximate optimization

algorithm

so these are what are called as the

variational algorithms and that started

the era of variational algorithms

while these algorithms were proposed in

2014 it was really in 2016 when ibm put

out its machine on cloud that it really

took a life of its own

then a lot of people started programming

because these are in tune with the

hardware limitations of the time and so

these algorithms were focused on

shorter depth you don't want to have a

deep circuit

that is not simulatable that is not

computable with the quantum hardware of

the current time in the niskara you

would want it more compact and you want

it in a way where

these are

hybrid algorithms as well that works

well in the cloud platform and also in

the kind of hardware regime that we are

in

um while the algorithms um and different

derivatives of these algorithms came

about a lot of applications are also

looked at

real world problems that then took these

algorithms and mapped onto some real

world applications i'm going to comment

on some of that in the later some of it

is listed quantum machine learning

pricing option com battery optimization

and so on so forth

but really

what is the hardware part so it's

important to understand

the background to all of this

this chart here shows the

lifetime evolution

of superconducting qubits

by lifetime what is meant by that is

that if you encode some quantum property

into a qubit how long can it sustain be

after which noise starts dominating the

system

uh on the x-axis is time starting from

about 2000 all the way till 2020

and on the y-axis is lifetime in

microseconds

and so

what you are seeing is the scale

starting from

nanosecond at the bottom so if you saw

if you see cooper pair box uh in around

uh

year 2000 they were hovering in the few

nanoseconds that is a quantum property

was visible was available only for few

nanoseconds

that was not very interesting but it was

an important

movement in the experimental side in

proving certain properties what the

specific implementations are are not

relevant but what is important to notice

is the trend

so if you see

the trend

it is been increasing linearly in this

plot however note that the y axis is in

log scale which means that the lifetime

has been increasingly exponentially in

the last 15 20 years and more so in the

last five to eight years it has been

much more rapid lot of activity has

happened because it's gotten more and

more real

notice that we are in hovering around

100 microseconds ish

somewhere here in the regime that we are

in

i will show some actual data from some

of the hardware and we are moving

further and further up the trend seems

to be that we are increasing the

lifetime and this is important to

understand because lifetime while we are

increasing exponentially it is still

limited so 100 microsecond for example

is the time that we have to do all our

quantum compute before noise takes over

so which means that all the calculations

that we have to do the algorithm needs

to be structured and be more compact and

more we

should be more shallow so that we fix or

compute all the things needed within

that fixed time budget that we have

recently um jakam better put out a tweet

indicating that

now we

have entered a regime of milliseconds so

it's an order of magnitude up so we are

moving from microsecond to a millisecond

second regime

um so

we are now having uh qubits that can be

in the order of milliseconds and more um

these are experimental at this point in

time soon hopefully we will have this

productized available in cloud but so

clearly we are seeing a trend that is

growing in a longer duration where

quantum properties can exist so that

means that we can now can do more and

more computation using this hardware and

the trend seems to be exponentially

increasing

so then

why

uh

quantum variational algorithms so this

is important to notice so these are

snapshots from actual hardware that we

have on cloud these are little bit of on

the higher end in terms of its fidelity

at this point this is as of 10th july

2021

so if you go to this particular link

where we have all the machines listed

you can go click on it and you will get

a picture something like this i've

chosen some subset of them uh ibmq

mumbai is the name of a particular

machine kolkata is another one montreal

is another one please notice that the

quantum volume of these are 128 which is

at this point in time on the higher side

and the important part that i want to

highlight is in the orange box here

what this

calibration data shows is the average

error so we can see that the c naught

error hovers around

10 power minus 2

roughly

readout error is little worse it's about

10 power minus 1

and then

average lifetime in this particular case

in mumbai uh is about 120 microseconds

so the c naught error is

roughly um

0.1 percent and readout error is

hovering around one percent error

and if you look at kolkata it is similar

to the mumbai machine

the

c naught and the average readout

average readout in kolkata seems to be

better but the average c naught error

hovers are on the same ballpark as what

we see in mumbai

montreal is little older

and you can see that

the c naught error is worse and also

the um the readout error is also worse

compared to the other machines

as you can see the lifetime is also

about 100 microseconds if you take

the mean of these two numbers

so you can see that

the current state of the art as we see

it hovers around 100 120 microsecond

lifetime c naught error is

readout error is dominating that is the

measurement error around one percentage

c naught error is about an order of

magnitude less but still is dominant as

the number of c naughts increase you

will have more and more errors in the

system and finally the measurement which

is the readout error causes lot of

noise in the system as well so what this

means is that this is putting constraint

on the algorithms that we have to come

up with which means that we have to have

shorter circuits to perform the

algorithm and we should be able to solve

a problem

of consequence which means that the

classical hardware should find it hard

to solve so you would want to solve

something that only a quantum

system can solve in a compact fashion

that the classical hardware cannot solve

only then we will have something of a

quantum advantage

using these platforms so the variation

algorithms forms into this falls into

this nice form factor that fits into

this limitations of the current hardware

which means that the variation

algorithms are structured in a way um

that it's a hybrid algorithm we're going

to see that in detail shortly

where you have a classical component

that is you're going to run aspect of

optimization in the classical hardware

but the key element of some of the

harder part of

the optimization for example calculating

the energy value or an expectation of

the lowest energy for example those are

com computed in quantum and then the

tuning part happens in classical so you

will see that in detail this is how the

variational structure comes in and that

sort of fits into the current

limitations of the hardware that is out

there

that is the reason why variational

algorithms or variational quantum

algorithms have gained a lot of traction

and particularly in the niskira

this is here to say

stay and more and more variations or

derivatives of these algorithms are

coming about and many applications

are being looked at using these

techniques so it's important to

understand what these are how it is done

what are the concept behind it and what

are the challenges therein and what are

the potential applications this will be

the topic of the next section