Quantum signature kernels

Arising from rough path theory, the signature transform captures features of time-series data by constructing a so-called path signature. This feature has proven valuable for various machine learning tasks. However, computing the associated signature kernel classically remains computationally intensive. In this talk, I will present recent developments in generalising the signature kernel to randomised Lie group path developments. I will discuss a quantum approach via matrix models with an associated unitary quantum signature kernel to propose a quantum algorithm for its computation.