Machine Learning in Quantum and Classical Domain

Yash Raj Saxena
4 min readOct 30, 2019

Back when the visible land was the only safe place known to humans, someone decided to go beyond the horizon in search for more land. Flying and soaring high up in the blue skies were only a mere dream for generations until someone decided they had to fly. Watching and talking to the moon was like talking to celestial beings, we can never be, then someone decided to talk and walk with moon hand in hand. Everything that humans have achieved so far was once a dream and all the more unachievable UNTIL someone decided to achieve it. A few decades ago smartphones were thought of as the technology that will change the future, which they did. But are they enough?

Human curiosity and desire to have more than we already do are what drives us to work hard every day. Now that as a race we have reached the acme of technological advances we thrive for more. What classical computing did for us over the last decade can be improved ten times over with the advent of quantum computing. Even though it is a fairly new field of work, its roots go back to the 1920s. Now when we are able to comprehend its capabilities to the fullest, we know in what direction to progress. This combination of direction and human desire is enough for us to reach a new age of technology and expert systems. What machine learning has done for us in the last 30 years, quantum machine learning will equivalently do in the next 5 to 10 years. The algorithms that today take days to train with tons of computational power will probably be completed in a matter of seconds if not less. Weeks and months of time wasted to train and improve machine learning models will be utilized to make further improvements to things of greater importance at that time. And I believe that it is my responsibility as a fellow researcher to help in this race to the future. My work helps my peers and other similar computer science students easily understand and grasp the “complex” quantum world.

Quantum Machine Learning although being actively researched has yet not reached thousands of researchers who can together bring about change faster than it already is coming around. The primary issue that affects most is the lack of knowledge in the field of physics, namely quantum mechanics. This stops many from actively becoming a part of this upcoming field. I hope that my work in bridging the gap between quantum and classical machine learning can help many more to step up and bring about significant change.

My motivation for taking up this project was a nexus of several ideas and beliefs. First and foremost was my dedication and dream of making the world a better place in the short human lifespan. This has been my dream for a long time and I believe I am completely capable of achieving it. Second was my love of machine learning and the zeal to study new algorithms. Third was also a big factor for me in taking up this project which was my level of understanding of the physical world and fundamentals of physics. All these factors together played a major role in motivating me to take up this topic as one of the most important parts of my educational career, my capstone project.

The aim of this project is to provide an easy and understandable documentation for new minds that wish to venture into the field of quantum computation and quantum machine learning. Even though there is a lot of documentation already available for people to access, learn and refer from, most of this documentation is either too complex for someone coming from a different background or very poorly written and compiled making it incomprehensible. It is because of this lack of documentation that many new researchers get discouraged and are not able to showcase their abilities.

Keeping these things in mind I planned out my project and aim to provide as much knowledge as I can in the most comprehensible way possible. The first step towards the completion of this thesis was me developing my understanding of quantum mechanics and quantum computing. I had a good grasp of basic under graduate level physics which helped me further learn more without much hassle. The first task was understanding how quantum particles and qubits work in order to understand quantum computing. Then come basic mathematical calculations for quantum algorithms like Shor’s algorithm and Grover’s algorithm. After these basics have been taken care of only then can one start getting involved with quantum gates, which will eventually become the base building blocks of quantum machine learning algorithms. Once an apt understanding of one, two and three qubit quantum gates have been developed along with mathematical practice of the same, only then can a person start quantum circuit building and understanding. This is when I worked on understanding and decoding quantum circuits for machine learning algorithms.

The machine learning algorithms I chose were Support Vector Machines (SVM) and Perceptrons. Although these algorithms have been developed by several individuals and organizations, each interpretation of them is different from the other. That is the crux of quantum computing. Each qubit is different for every quantum computer in the world. Several organizations are working round the clock to develop these state-of-the-art systems but each one of them is fundamentally different from the other as each qubit has a different quality.