IBM’s New Tool Lets Developers Add Quantum Computing Power to Machine Learning
IBM is launching a new module as part of its open source quantum software development kit, Qiskit, to enable developers to leverage the capabilities of quantum computers to improve the quality of their machine learning models.
Qiskit Machine Learning is now available and includes the computational building blocks needed to integrate machine learning models into quantum space.
Machine learning is a branch of artificial intelligence that is now widely used in almost all industries. The technology is able to browse through ever-larger datasets to identify patterns and relationships, and ultimately discover the best way to calculate an answer to a given problem.
SEE: Green energy policy of computer data center (TechRepublic Premium)
Researchers and developers therefore want to ensure that the software delivers the most optimal model possible – which means increasing the quantity and improving the quality of the training data fed to the machine learning software. This process inevitably results in higher costs and much longer training times.
Delegating parts of the process to a quantum computer could solve these problems, speeding up the time it takes to train or evaluate a machine learning model, but also dramatically increasing what is known as the feature space – the whole features that are used to characterize the data provided to the model, for example “gender” or “age” if the system is trained to recognize patterns on people.
While classical computers are limited by the computing power required by large feature spaces, quantum computers should – once the technology matures enough – excel at performing large calculations in a short time.
With quantum computing still in its infancy, much of the work around quantum machine learning is theoretical and still depends on scaling quantum devices in the future; but a growing number of researchers are nonetheless showing an interest in exploring the opportunities that technology could one day open up.
“Quantum computing offers another potential avenue to increase the power of machine learning models, and the corresponding literature is growing at an incredible rate,” said the Qiskit applications team. “Quantum machine learning offers new types of models that exploit the unique capabilities of quantum computers to, for example, work in exponentially higher dimensional feature spaces to improve model accuracy.
“The use of classical and quantum machine learning models can give researchers a better understanding of quantum chemistry and physics, opening up many new applications and research directions.”
However, even for the most savvy machine learning developer, stepping into the world of quantum can be a daunting prospect – which is why Qiskit released the new module, with the promise that the program’s design allows developers to prototype a model even without expert knowledge. of quantum computing.
For example, Qiskit Machine Learning provides QuantumKernel, a tool that calculates kernel matrices for a given set of data in a quantum framework. This is the first step towards mapping data into an exponentially higher dimensional feature space that can provide more precise training for machine learning models.
The new module also contains several implementations of quantum neural networks, along with learning algorithms to train and use them, so developers can build and test their own networks.
Finally, Qiskit Machine Learning allows users to integrate their new quantum neural networks directly into the open-source machine learning library PyTorch. A platform developed by Facebook, the PyTorch library is primarily used for applications such as computer vision and natural language processing.
Indeed, as Qiskit’s applications team explained, quantum machine learning is expected to work in tandem with classical computing, with computational heavy tasks performed on quantum devices to improve models designed for them. classic applications.
“They can be part of a more complex computation, such as a deep neural network that consists of classical and quantum layers,” the team said. “This opens up endless opportunities to study the potential power of quantum neural networks for a large number of applications.”
Once they build a quantum machine learning model in Qiskit, developers will be able to test the algorithm on classic computers, but also on IBM’s cloud-based quantum systems. The first release of Qiskit Machine Learning provides a starting selection of models, but because the platform is an open-source library, the applications team has encouraged researchers and developers to get down to work to start expanding. the body of research.