Machine learning with quantum computers

Machine learning is considered a key technology of artifical intelligence (AI). Machine learning methods learn connections from large amounts of data that would remain hidden from our human brain due to their complexity. These machine learning methods differ greatly in their approach from how our human brain works. They depend on a huge number of examples for successful pattern learning.

In the case of complex relationships, correspondingly complex models help to describe the relationships. This results in complex computing processes that are cuurently solved on special hardware such as GPUs (Graphic Processing Units), TPUs (Tensor Processing Units) or FPGAs (Field Programmable Gate Arrays). Depending on the area of application, these technologies are already reaching their limits. We want to overcome these limits and also understand the connections between increasingly complex systems in the future.

With quantum-assisted machine learning, a new era of possibilities starts here. On the one hand, the training time can be greatly reduced, on the other hand, the accuracy of the procedures can also increase.

Algorithms in use

Classification	Neural Netwprk
Among other things, support vector machines are used in the classification of data. If the data cannot be separated with a simple straight line, so-called kernels are used, which map the data in higher-dimensional spaces. Quantum Support Vector machines can use new types of kernels due to the properties of quantum bits. This can increase the prediction accuracy of the model.	Neural networks are used today for complex machine learning tasks enabling regressions or classification. With a quantum neural network, parts or all layers of the network are replaced with quantum layers. Instead of classical weights, rotation angles of qubits are now learned. Through this, a faster learning process and better convergence behavior can be achieved.

Classification

Neural Netwprk

Among other things, support vector machines are used in the classification of data. If the data cannot be separated with a simple straight line, so-called kernels are used, which map the data in higher-dimensional spaces.

Quantum Support Vector machines can use new types of kernels due to the properties of quantum bits. This can increase the prediction accuracy of the model.

Neural networks are used today for complex machine learning tasks enabling regressions or classification.

With a quantum neural network, parts or all layers of the network are replaced with quantum layers. Instead of classical weights, rotation angles of qubits are now learned. Through this, a faster learning process and better convergence behavior can be achieved.