By Amrita Manzari, QSI Head of AI & Quantum Machine Learning
In 1935, Einstein wrote a paper with Boris Podolsky and Nathen Rosen trying to expose the weird behavior of quantum mechanics calling it “spooky action at a distance”. Among its many weird behaviors, the notion of quantum superposition really defies our imagination.
Even weirder, if one looks into a quantum system of say two electrons and they are in an entangled state, if you measure the property of one electron, say its rotation, you can tell what the other electron’s rotation is — without even bothering to measure it. Is it weird or astonishing? I think both.
A lot of these complex concepts seem very difficult to understand at first, it is mind-boggling for a Software Engineer who works in classical computer logics and data structures. As a technologist with a background in machine learning, software engineering, and data science, it takes myriads of trials, experimentation, and perseverance to get the concepts behind the mechanics of quantum machine learning. But I think it’s always better to experiment, especially in the quantum world.
The idea is simple: “Start slow and then take a pace”.
Let’s take a pause and understand what is Machine learning and why we should bother about quantum machine learning?
Machine learning is a subset of Artificial Intelligence where a machine takes the data, performs computations to recognize patterns and learn from the data to output correct predictions. It performs computations using mathematical models of data to help a computer learn without direct instruction. This enables a computer system to continue learning and improving on its own, based on experience. Machine learning allows a machine to automatically learn from past data without programming explicitly.
The goal of machine learning is to allow machines to learn from data so that they can give accurate output. In AI, we make intelligent systems to perform any task like a human. In machine learning, we teach machines with data to perform a particular task and give an accurate result.
Quantum Machine Learning is an integrated space of quantum algorithms, quantum computation, and machine learning.
Quantum computers use the concepts of quantum coherence, superposition, and entanglement to process information in such a way that classical computing cannot do. The ability of quantum states to be in superposition can lead to a substantial speedup of a computation in terms of complexity since operations can be executed on many states at the same time.
A quantum algorithm is the stepwise procedure performed on a quantum computer to solve a problem, for example- searching a database, factorization of large numbers, and optimization with the latter effectively used in a speed-up of machine learning algorithms. In quantum machine learning, quantum algorithms are developed to solve typical sort of problems of machine learning using the concepts and efficiency of quantum computing. This is usually done by adapting classical algorithms or their expensive subroutines to run on a quantum computer.
So, basically, we can grab two concepts here, ML techniques can be used to output the quantum processes. Also, quantum computational concepts can be used to design and enhance ML algorithms.
“Q-CTRL scientists developed a new way to process the measurement results using custom machine-learning algorithms. In combination with Q-CTRL’s existing quantum control techniques, the researchers were also able to minimize the impact of background interference in the process. This allowed easy discrimination between “real” noise sources that could be fixed and phantom artefacts of the measurements themselves.”Source: The University of Sydney on Q-CTRL
Now, the question arises “why should a Machine Learning expert be interested in Quantum Computation?”
As Frank Zickert argues in his book, “Hands on Quantum Machine Learning with Python”, “Quantum Machine Learning promises to be disruptive. It requires a new set of developers. Developers who understand machine learning and quantum computing. Developers capable of solving practical problems that have not been solved before. A rare type of developer. The ability to solve quantum machine learning problems already sets you apart from all others. Don’t miss it!”
In thinking about the potential impact of Quantum Machine Learning, I include the points below:
- The proximity to the physical limits of classical computing paradigm restrains the efficiency of classical Machine Learning, along with the increasing number of datasets.
- Hard Classical problems might benefit from Quantum Computational Paradigm
- Classical Machine learning algorithms can be used to support Quantum processes for e.g.- Quantum Error Correction
This implies that hard classical problems might benefit significantly from the adoption of quantum-based computational paradigms. Also, classical machine learning algorithms can benefit quantum computation in different processes. So, the best way to understand the concept of quantum machine learning is as a unified and hybrid field that combines advantages of both classical machine learning and quantum computation.
How to transition from Software Engineering or ML Engineer to Quantum Machine Learning?
It has now almost been eight years; I have been working in software engineering and machine learning domain. One thing that is quite observable is, ‘the extensive amount of computation when data points are projected in high dimensions during machine learning tasks makes it hard for classical computers to manage and compute’. This inspired me to dig deeper into the computational domain of solving hard problems. The finance and business domains have several instances of data computation and representation where we observe a lot of computation, whether in data modelling, statistical representation, or optimization. Even if we are using tensor flow optimization, it is hard for classical computers to deal with such large computations.
The world of quantum computing is not just full of esoteric concepts, searching for right candidate who can work on these concepts is even more difficult. Employers need candidates with interdisciplinary skills including data science, engineering, physics, and computer programming. The most common level of specialization is PhD. in mathematics, physics, engineering, artificial intelligence, and related fields.
However, the increased requirement of quantum engineers, especially in the quantum machine learning area, has led to several training programs for professionals and technologists, software Engineers and C-Suite people. Further, while the available training programs work on very high levels of abstraction, we need people who understand intricacies with great expertise.
Although quantum machine learning is a relatively new area and requires specialized skills, we need individuals who are extraordinary and possess backgrounds from both the quantum and the software domains. It is quite a challenge to find individuals who can map problems to quantum space, sometimes just holding a PhD. in physics and mathematics might not be enough. It requires expertise in data modelling, data analysis, engineering, as well. And finding the right mix in a curriculum vitae (cv) is certainly rare.
With the right mix of passion and curiosity to understand the intricacies of how a quantum system works, also how it can be applied to real world problems, I think people can transition to QML roles.
Based on my research from different quantum job requirements and interviewing various experts from quantum machine learning area, I am putting down below a few parameters that are required to transition to quantum machine learning. The below chart illustrates the primary skills that are required in the areas of educational background, technical skills, business skills and soft skills.
Image Source: Quantum Strategy Institute, 2022
Source: Quantum Strategy Institute, 2022
A quick glance at the comparison view between Quantum Machine Learning Engineer and Quantum Engineer. For more details on Quantum Engineering, please refer to the article: Roadmap to Quantum Engineering.
Let us discuss each section in more detail.
- Masters/PhD. in Quantum Computing, Science (Physics/Chemistry) or Engineering- Masters/ PhD. in quantum computing, machine learning, artificial intelligence, physics, mathematics, chemistry, or related fields are an ideal fit. However, organizations like United Health Group, JP Morgan, Standard Chartered, Accenture, HSBC, etc. are working on designing training programs with the focus on software engineers, technologists, and C Suites. The industry has started realizing the impact of quantum with a primary focus on transportation, space, life sciences, insurance and healthcare sectors. Since the field of quantum machine learning is still in its budding phase and, therefore, PhD roles are an ideal fit for the organization to pursue research. However, we will see a lot of dynamics in terms of Industry roles coming in the span of 5-10 years.
- Machine Learning Engineer/AI Engineer – Machine Learning /AI is one of the important skills for quantum machine learning. However, for an AI/ML Engineer, exposure to and understanding of quantum theory and computation is very much required.
- Research Associate/Scientist– A research associate or scientist already working in quantum computation, optimization and AI/ML is a plus. Such experience can bring great expertise in this domain.
- Internships and Certifications– There are a lot of internships and certification programs available. However, one should know, not all internships and certification courses provide in-depth knowledge of quantum computing or machine learning. You need to self-invest a lot to acquire depth in both subjects.
- Industry Experience – Experience and knowledge gained from the commercial industry is a great addition. Understanding of the business and consumer helps in determining and deriving several use cases of a particular Industry front.
Talking to academic and industry experts on what technical skills could define a quantum machine learning engineer, Adrian Perez Salinas, Research Engineer and pursuing PhD at Barcelona Supercomputing Center, casts some light on the technical requirements for QML Engineer and Quantum Computing in general. As per some of the pointers that I grabbed from our discussion and my research:
- AI/ Machine Learning– Quantum AI/ML needs to achieve several milestones to become mature in the field of quantum engineering. Widespread adoption of open-source modelling and training frameworks is one critical point. Before starting with quantum machine learning, one should know the AI and ML algorithms in depth.
- Applied Mathematics/ Linear Algebra– Linear algebra, complex numbers, probability theory are required skillsets.
- Quantum Computation and Quantum Information– Harnessing quantum physics for the purpose of computation holds significant promise for the development of data science and machine learning. Also, in turn, data science and especially machine learning can play a major role in the development of quantum computation and quantum information.
- Quantum Machine Learning Algorithms– Understanding of quantum machine learning algorithms and how they can work with classical machine learning algorithms. Some examples include:
- Quantum Machine Learning Algorithms for supervised and Unsupervised Learning
- Quantum Principal Component Analysis
- Quantum Support Vector Machines and Kernel Methods
- Optimization Techniques
Refer to the graph below to check more quantum machine learning algorithms:
Source: Quantum Strategy Institute, 2022
- Quantum Optimization and Simulation– Quantum computers have a natural tendency to simulate the dynamics of quantum systems. Also, algorithms such as classical discrete optimization problem becomes an eigen problem on a large matrix, that can be visualized with quantum systems. Understanding of both quantum optimization and simulation is a plus.
- NISQ– We are still amid what we call as “Noisy Intermediate Scale System”. The noisy intermediate scale quantum devices (NISQ) are being constructed out of cold atoms, superconducting quantum circuits, trapped ions, and other quantum systems for which we have achieved an exquisite degree of control. Some of the tasks for these NISQ devices will be as simulators (or emulators) of other highly entangled quantum many-body systems. The goal is to supplant our current conventional computer simulation technology, such as exact diagonalization, quantum Monte Carlo, or tensor network methods.
- Tensor Networks– “Tensor Networks are extremely powerful because they represent huge amount of data, so it is a way to represent large amounts of data in a very efficient way. We think that it would be like the linear algebra of 21st century and maybe you want to know that because there are some extremely useful methodologies that can be used for processing images, processing sound.” says Adrian Perez Salinas, Research Engineer and pursuing PhD at Barcelona Supercomputing Center.
The science of quantum machine learning is complicated. For success in this field, along with strong educational and technical background, there are a few other parameters that are required as well:
- Right Mentorship: Finding the right mentorship is very important. Few initiatives, such as the Quantum Open-Source Foundation (QOSF) programs, are entirely designed in the framework of Mentee- Mentors that helps in connecting with the quantum enthusiasts and right mentors.
- Contribution to Scientific community– Contributing to the scientific community by participating in hackathons, slack channels, and networking helps in finding the right people and a great value add to the knowledge content. In order to enhance your networking skills, there is this interesting website to check: http://quantumapalooza.com/, which aggregates a list of all the free quantum computing meetup events going on globally.
- Contribution to research papers– Writing research papers, original content, or maintaining a GitHub account with your coding projects shows the level of depth. It is not, however, a mandate for certain roles. But it is always an added advantage, especially if a person is seeking for scientific and research domains.
- Understanding the Business and Industry requirements- In the commercial world, it doesn’t matter how cool your tech is if it’s not getting to the customer. Things like timescales, hard deadlines, user experience, economic cycles, quantum cycles (Spring, winter, managing hype etc.), message to market, product creation etc. are all critical skills that the workforce as a whole needs to know. Business development and Quantum Research Architects are some of the challenging roles in the quantum machine learning area.
- Understanding the Consumer– Understanding the consumers requires more outlook into what type of market or industry you are selling your products to. Experience Manager roles aare some of the challenging roles as they have daily interactions with the customers to understand their requirements and also how they want to see the software packages.
Soft skills play a major role in enhancing the path to development. Continuous development, strong determination, perseverance, curiosity are very important to evolve and grow. In the case of quantum machine learning, shaping the right mindset and setting the right expectation plays a major role.
Teamwork, collaboration and accountability are some major skills to succeed in a project and Team environment. Quantum machine learning is still young and evolving. “There’s a famous quote about lasers, when they were first discovered, that said they were a solution in search of a problem. Now lasers are used everywhere. Similarly, we suspect that when quantum data will become highly available, quantum machine learning will take off.” , Patrick Coles, a quantum physicist at Los Alamos National Laboratory, “Is Quantum Computing the Future of AI?“.
A word of caution, there is a lot of potential in Quantum AI/ML, but it is still in the process of realization. We might get immediate results, or we may have to wait for this field to mature. But the important aspect is to set the right expectations and not to get impatient if we do not get results immediately, like in classical machine learning. All we need to do is to work and contribute for breakthroughs that are much expected to happen, especially in the quantum machine learning area.
Quantum machine learning is an intense science. The intersection of multi-disciplinary areas makes it more complex both in theoretical and practical terms. There is optimism towards the advancement of quantum machine learning field, all we need to work towards is realizing the potential it could bring to AI and ML domain.
I hope this helps you in your journey of transitioning to quantum machine learning. Stay tuned to The Quantum Strategy Institute for the next release in quantum machine learning.
The Quantum Strategy Institute’s purpose is to demystify quantum technology and encourage the development of a pragmatic quantum mindset within the global business community, sharing practical applications and offering strategies for its successful adoption.
Amrita Manzari is Head of Artificial Intelligence and Quantum Machine Learning at the Quantum Strategy Institute. She is also Associate Manager Software Engineering at United Health Group and a Quantum Enthusiast. Amrita is based in India.
A special mention to:
- Dr. Frank Zickert, Author of Hands-On Quantum Machine Learning with Python
- Adrian Perez Salinas, Research Engineer and pursuing PhD at Barcelona Supercomputing Center
 Hands- on Quantum Machine Learning- Volume 1, Dr. Frank Zickert
 Quantum Computation and Quantum Information, Michael A Neilson, and Isaac L. Chuang
Copyright 2022, Quantum Strategy Institute