Quantum machine learning (QML) combines machine learning concepts with quantum computing. Quantum computing is viewed in many ways as a complementary accelerator, enabling hybrid quantum-classical machine learning workflows — subsequently, quantum machine learning is being researched and applied to classical machine learning models to help achieve better results, in less time, and at less cost than in the past.
Machine learning experts can recognize two potential benefits of quantum machine learning over the traditional classical possibilities.
For starters, QML models can increase learning efficiency because they have the potential to learn from smaller amounts of data than their classical counterparts. This means QML could have use cases to improve classical models in areas with large amounts of complex data such as climate research, pharmaceuticals, logistics, and more.
In addition, QML models can reveal more correlations in a data set, due to their ability to run through far more permutations and analyze the data yielded from each interaction. This in turn helps allow for predictive models to improve their ability to extrapolate outcomes from a given set of situational variables.
Potential Real-world Quantum Machine Learning Uses
There are potentially promising applications of QML that use optimization to solve problems.
Perhaps one of the most famous of these problems is the ‘traveling salesman’ problem. Here, quantum computing will aid in finding solutions by being able to consider many paths at once, as opposed to the serial method which makes such problems intractable for classical computers.
In addition, the combinatorial optimizations that a quantum computer processes can adopt similar optimization strategies for finding the best parameters in machine learning models. One of the most promising current potential applications of this is for classifying images. The scale of today’s quantum computers makes them already very effective for correlation analysis for small data sets, such as the kind in radiography images.
Overall, analysts estimate that the potential benefits of quantum computers could unlock as much as $140b in value from solving QML problems in automotive, finance, and tech.
Below, we look at one of the exciting applications for quantum machine learning that IonQ has already worked on with a current customer: autonomous driving.
Case study: Improved Image Classification through Quantum Machine Learning
Image classification is pervasive in our everyday lives, from the photo apps on our phones to the systems that track the packages dropped at our doorsteps. However, this is not a solved problem, especially when image recognition is helping control a moving vehicle. A challenge is that several steps in the most common machine learning approaches for image classification are very resource-intensive. In a world with plentiful and inexpensive classical bits, the solution has been to throw more compute resources at the problem.
In the video below, we share the story of how one forward-looking auto manufacturer approached this problem, performing function optimization to minimize the difference between their model's variables and the ideal values for a self-driving car to 'recognize' road signs and react appropriately.
Learn Quantum: Machine Learning Image Recognition Application
Explainer video on IonQ's recent application project with Hyundai
To see more videos in IonQ’s ‘Learn Quantum Video Explainer Series’ go here.