Machine learning (ML) and quantum computing (QC) are some of the hottest topics at this moment but they are rarely combined into one field of study. QC completely revolutionize the way we think of computing with theoretical results that tremendously speed up many algorithms, and ML has significantly improved the various artificial intelligence (AI) applications especially in neuroscience and hearing systems. But how can we leverage quantum computers to accelerate this development in AI even further? Only few companies and research groups focuses on this interface, but those who do predicts a tremendous increase in both speed and capacity of these learning algorithms that can enable us to learn far more complex structures in some cases exponentially faster than today.
One of the biggest challenges in QC lies within the engineering of the hardware and it may be many years before we build robust and scalable devices. However, there exists a subset of quantum technologies known as “near-term” quantum computers, which we expect to be able to program and run within few years. These small scale quantum information processors can produce statistically patterns which remain hard to produce with classical computer resources. The hope is that near-term devices enables us to recognize statistically complex patterns too.
This project aims towards deriving and developing quantum algorithms to solve computationally hard problems within ML with a starting point in the inference of generative models on near term devices as well as continuous variable processors. More lately, results in deep learning has provided us tremendous capabilities in data science with generative adversarial networks (GANs) and variational autoencoders (VAEs) being some of the most popular ones. These architectures allows us to learn very complex structures in high dimensional data and subsequently sample new artificial data from these distributions. We therefore hope to push the frontier by implementing and running QML algorithms on state of the art quantum devices and simulators in order to apply, test and benchmark theoretical quantum results against classical counterparts. We wish to develop generic tools for computer scientists to apply quantum machine learning, which facilitates an easy-to-access framework for increasing capacity, complexity and speed in learning algorithms.