Special ACM Seminar
Maud Lemercier received her doctoral degree in Statistics at the University of Warwick as part of the Oxford-Warwick Statistics programme (2018-2022). Prior to her doctorate, she obtained a Diplome d'Ingenieur from IMT Atlantique and an MSc in Machine Learning from Imperial College London. She is currently a postdoctoral researcher at the Mathematical Institute of the University of Oxford, and a visiting researcher at the Alan Turing Institute. Her primary research area lies in the theory of rough paths and its applications in addressing high-dimensional statistical learning challenges. Within this context, her recent research primarily focuses on developing kernel methods and anomaly detection techniques tailored to sequential data. She is passionate about exploring how these approaches can contribute to diverse scientific disciplines such as biology and radioastronomy.
Neural stochastic differential equations (Neural SDEs) have proven to be powerful continuous-time generative models that leverage neural networks to parameterize the drift and diffusion functions of stochastic differential equations (SDEs). These models have achieved state-of-the-art performance in generating multivariate time series through adversarial training as GANs. In this talk, I will introduce Neural stochastic partial differential equations (Neural SPDEs), which extend the capabilities of Neural SDEs to model spatio-temporal dynamics in continuous space-time. I will start by explaining how the Neural SPDE model parametrizes the mild solution of an SPDE and then discuss how it can be trained by minimizing suitable scoring rules on path space based on signature kernels.