A Paradigm Shift in Quantum Machine Learning

Quantum computing is a disruptive computing technology that has the potential to revolutionize machine learning. At present, almost all quantum machine learning (QML) approaches focus on inferencing techniques, for e.g., variational quantum algorithms, QAOA, quantum neural networks, and quantum support vector machines. These quantum inferencing techniques are primarily applicable for analyzing data that is inherently quantum. However, our current machine learning infrastructure is largely classical—the data produced is classical, the training algorithm is classical, and the inferencing hardware is also classical. We almost never produce or deal with quantum data within the current machine learning infrastructure. Therefore, the applicability of the current quantum inferencing techniques is extremely limited, probably only to the quantum physical experiments that produce quantum data. To realize the full potential of quantum computers for machine learning applications, we need a paradigm shift in quantum machine learning. We must use quantum computers for training machine learning models as opposed to using them for inferencing. In this talk, I will present the results that we obtained by using quantum computers to train machine learning models. Specifically, I will present the results we obtained when we used adiabatic quantum computers to train widely used machine learning models such as linear regression, support vector machine, and k-means clustering.