Quantum Machine Learning- A Beginner’s Guide
An entry portal to the world of QML
Does a machine learn?
Have you ever seen a robot movie, for example, the 2010 Tamil blockbuster film Enthiran, where the robot thinks, expresses, and almost behaves like a human, and wondered whether science can advance so much and achieve this feat? Well, welcome to the world of Machine Learning, where you can train a machine, to analyze, work, and behave like a human, just like how your train a baby or a pet dog.
Machine Learning, just like how the title suggests is all about how you make a machine learn, and make it think independently. Just like how we have a varied curriculum of education like CBSE, ICSE, State Board, etc., similarly, we have a variety of techniques to train the machine or technically called, machine learning algorithms. Let us explore a few of the prominent machine learning algorithms.
Any machine learning algorithm falls under one of these three categories, namely Supervised Learning, Unsupervised Learning, and Reinforcement or Semi-Supervised Learning.
1. Supervised Learning
A teacher feeds in the data from a textbook or a reference material to the student. Thus, the teacher knows and supervises the student. Similarly, a monitored set of data is fed into the machine, and the learning pattern is supervised. This is called supervised learning. Here, the input variable, assume x, is mapped to the output variable, assume y, as y = f(x) where y can be predicted for any x in the future.
Some of the examples are Linear Regression algorithm, Random Forest, and Support Vector Machines.
2. Unsupervised Learning
This is an independent study type. The student is given a set of data and he has to gain experience through his own creative way of studying and classifying the data. Similarly, a random data set is handed over to the machine and it has to come up with its own way of classifying it. In short, the machine has to find on its own y for the given x.
Some algorithms under this classification are k-means and aPriori algorithm.
3. Reinforcement or Semi-Supervised Learning
This is a mix of both of the above ways. Here a student is given a few examples and is expected to face the rest of the questions concerning those examples. A machine here is given a set of data that is classified and is given a larger set of unclassified data. It is expected to classify the data with the help of previous inputs and outputs. This is more preferred than the above methods since the involvement of experts is minimal and the expense and time involved is lesser than supervised learning. The machine learns through the trial-and-error method
Markov Decision Process is one of the algorithms that encompass this method.
The above organizer chart(Fig 1) shows the various algorithms used in machine learning
Bits and Bytes, the Qubit way
Let's first learn a bit about quantum computing before we get to know what quantum machine learning is all about.
Quantum computing refers to algorithms that use certain algebraic methods like linear algebra, or vector algebra on a conceptual computer called the quantum computer. Some of the prominent quantum computing models are quantum circuit model, quantum Turing machine, adiabatic quantum computer, one-way quantum computer, and quantum cellular automaton (not automation)
Quantum computation is done in terms of quantum bits or qubits, which consists of 0s and 1s. Some quantum computers use logic gates in place of classical computation tools. Quantum computers follow Church-Turing thesis which states: “a function on the natural numbers can be calculated by an effective method if and only if it is computable by a Turing machine” and the Turing machine here refers to the quantum computer.
Quantum computers have a possible scope in many areas, but to date, quantum computers are just “hypothetical”. But the various mathematical methodologies involved are used and researched further to make quantum computers, or maybe quantum robots too!
ML is fine, but what is QML?
Now that we have an idea of what machine learning is, and what a quantum computer is, we now can fathom that quantum machine learning involves quantum computers and how they can be trained using machine learning
The above diagram(Fig 2) explains the difference between classical machine learning and quantum machine learning.
To know further about Quantum Machine Learning, check out our session ‘Quantum Machine Learning- 101’ by Mr. Santanu Ganguly, Lead Technologist, and Architect at CISCO, UK on 2nd of December 2020, at 5 pm IST.
