aqmlator package
Module contents
Machine learning (ML) is one of the fastest-progressing research directions in applied computer science. The field investigates the development of algorithms that can learn from data by fitting a collection of model parameters to the data by iteratively optimizing an objective function. The selection of a model structure, be it a neural network topology or kernel function, is a problem-dependent task often made by hand. But there exist Auto ML systems that can choose a model automatically depending solely on the input data and task.
Quantum computing (QC) studies how hard computational problems can be efficiently solved using quantum mechanics (QM). A large-scale error-corrected quantum computer can solve computational problems that don’t have a classical solution. A prime example of that is Shor’s algorithm for integer factorization. The ‘holy grail’ of applied QC is the so-called quantum supremacy or quantum advantage. That is the name for a technological milestone that would mark the moment when quantum machines will solve a specific task faster than the most advanced supercomputer. Although there have already been several quantum supremacy claims in recent years, there are no practical problems solved using quantum computing yet.
The search for such practical problems focuses on applications in the soft computing areas that are less susceptible to current quantum hardware imperfections. One of the possible applications of QC is quantum machine learning (QML). Quantum computers can be employed to build ML models that can be fit to data and then used during the inference process. In one of the QML scenarios, a variational quantum circuit forming a quantum neural network (QNN) constitutes only one part of the ML data processing pipeline. Since designing such a pipeline with a quantum component is challenging for non-experts in QC, we propose an auto ML solution that uses QML techniques.
We call our solution (A)uto (Q)uantum (M)achine (L)earning platform—AQMLator. It aims to enable data scientists, especially those without knowledge of QC, to use QML models. Given the data, the system will seek the model that fits the data best. Additionally, it will propose the circuit structure of the VQC model and its weights. Ultimately, the user can use the model as-is or encompass it in a hybrid model e.g. by extracting the proposed VQC model as a torch layer.