Software Between Quantum and Machine Learning

Building upon our toolbox QML-Essentials, we went a level deeper (pulling an old "Inception" joke here) and introduce, among some other shiny new features, a pulse-level simulation capability to this library. Specifically, we developed our own simulator from scratch (paper upcoming) which granted us with great flexibility and allowed us to build a comprehensive pulse-level simulator alongside a statevector and density matrix simulator. This new simulator is entirely built upon JAX allowing us to natively fully differentiate every computation in our simulation.

As you can see, there are many other things going on here, for example we introduced a topology-based approach of building Ansätze (a community term for describing the structure of a quantum model), which not only significantly reduced code complexity but also allows for research towards generative design of such Ansätze. Moreover we extended the set of available entanglement measures, added a generative (Fourier) training dataset and a first draft of an optimal control pipeline, used to tune the aforementioned pulse parameters.

We're proud that this paper got accepted to QSW26. Feel free to checkout a preprint at Arxiv.