Abstract of Edoardo’s talk// Despite a large, impactful body of literature on quantum machine learning (QML), many applications and methods remain uncharted, opening up prospects to observe an advantage in domain areas like chemistry. As in classical ML, optimally representing the input data in terms of quantum operators is key for QML; this is especially true with near-term quantum devices, where dephasing and other errors can hinder the scaling of the accessible solution space and the precision of these solutions. In this talk, Edoardo will share one of the group’s latest representation-learning approaches to encode molecular structure in a quantum circuit representation, and how this compares to alternatives. Through this example, he will then point out more general open problems in QML, like trainability, barren plateus, kernel concentration, and arguments of classical simulability, which may help visualise what’s next in QML.

Abstract of Choy’s talk// Several challenges continue to persist that attenuate the success of Variational Quantum Algorithms (VQA) in harnessing their true potential in obtaining good solutions to the problems they seek to solve. The most commonly cited problem is the emergence of barren plateaus as the size of the quantum circuit increases, and is particularly exacerbated by poor ansatz design and stochastic noise associated with imperfect qubits. However, a second intertwining layer of complexity arises from the increased proliferation of stationary points with poor solutions from an increase in the number of ansatz layers for underparameterised quantum circuits, and this phenomenon is relatively underexplored. In this talk, Choy seeks to explain the work that has been conducted aiming to broaden the repertoire of VQA previously explored with energy landscape theory, namely hardware-efficient ansätze and the quantum approximate optimization algorithm. Two key areas of expansion include the use of chemically-inspired ansätze in the electronic structure problem, and the use of quantum machine learning techniques in predicting chemical phenomena.