Quantum Computing and Machine Learning

Two of the most rapidly growing areas in computer science are machine learning and artificial intelligence. In simple terms, machine learning means training a computer to recognize patterns—either spatial patterns or temporal patterns or a combination of the two. More generally, machine learning is a procedure by which a computer can extract information from a database. The training consists of giving the computer some input data and the desired output and having the software work out connections that allow the computer to recognize similar patterns in input data on which it has not been trained. For example, you might give the computer several images of different kinds of dogs. The computer analyzes those images and comes up with some rules that allow it to produce an output “dog” when similar images are presented to it later. Much of machine learning involves solving systems of linear equations. We mentioned in the previous chapter that there are quantum algorithms that can carry out those solutions more efficiently than classical computers.

Related to image recognition is image processing. As a concrete example, cameras in satellites in Earth orbit send images that are useful for weather prediction, monitoring environmental changes (e.g., erosion and glacier deterioration), and military reconnaissance. If we could put fast image-processing computers on satellites, we could reduce substantially the data-transmission and reception requirements (Fan et al., 2017). Once QCs become more portable, image processing is likely to be one of the first space applications of QCs. Until then, quantum machine learning is likely to focus on solving problems related to quantum mechanics, such as quantum simulation.

Artificial intelligence (widely known as AI) is similar to machine learning. The computer is trained to sense its environment and learn what happens when it responds to that environment in various ways as it tries to achieve some specified goal. For example, AI allows a computer to learn to be competitive in strategic games such as go and chess, to control an autonomously operating automobile, to provide intelligent routing of messages in content delivery networks, and to run sophisticated military simulations. With AI, the computer is supposed to be able to learn all these things without human intervention.

What is the role of quantum computing in machine learning and AI? The complete answer is not yet in, but current work focuses on using quantum computer algorithms to speed up mathematical operations that are part of classical machine learning and AI. These operations include solving systems of linear equations, using what is called “principal-component analysis” to focus on the major features of large data sets, and finding eigenvectors and eigenvalues of matrices.