The motivation for this talk was to discuss effective data imputation strategies. Electronic Healthcare Records (EHRs) are inherently sparse in nature due to inconsistent sampling of parameters (like Weight, Hemoglobin level etc). Gaussian processes (GPs) are effective in making such data rich. This can have a strong effect on subsequent analyses (Sickness forecasting etc.). GPs have the utility of providing confidence bounds on prediction in addition to being more effective for smaller data. The original work by C. E. Rasmussen & C. K. I. Williams was summarized with the help of an intuitive guide by Jie Wang. 

Work by Murpy et al. Forecasting the evolution of contagion dynamics is still an open problem to which mechanistic models only offer a partial answer. To remain mathematically or computationally tractable, these models must rely on simplifying assumptions, thereby limiting the quantitative accuracy of their predictions and the complexity of the dynamics they can model. The paper proposes a complementary approach based on deep learning where effective local mechanisms governing a dynamic on a network are learned from time series data. Proposed GNN architecture makes very few assumptions about the dynamics, and they demonstrate its accuracy using different contagion dynamics of increasing complexity. By allowing simulations on arbitrary network structures, their approach makes it possible to explore the properties of the learned dynamics beyond the training data. Finally, they illustrate the applicability of our approach using real data of the COVID-19 outbreak in Spain. The results demonstrate how deep learning offers a new and complementary perspective to build effective models of contagion dynamics on networks.

Work by Guo et al. Graphs are important data representations for describing objects and their relationships, which appear in a wide diversity of real-world scenarios. As one critical problem in this area, graph generation considers learning the distributions of given graphs and generating more novel graphs. Owing to its wide range of applications, generative models for graphs have a rich history, which, however, are traditionally hand-crafted and only capable of modeling a few statistical properties of graphs. Recent advances in deep generative models for graph generation is an important step towards improving the fidelity of generated graphs and paves the way for new kinds of applications. This article provides an extensive overview of the literature in the field of deep generative models for graph generation.

Work by Gururajan et al. Poster presentation at the 39th ASI conference. Abstract is as follows : We apply the generalized Lomb-Scargle periodogram to the ^{123}I and ^{99m}Tc decay rate measurements based on data taken at the Bronson Methodist Hospital. The aim of this exercise was to carry out an independent search for sinusoidal modulation for these radio-nuclei (to complement the analysis in Borrello et al) at frequencies for which other radio-nuclei have shown periodicities. We do not find evidence for such a modulation at any frequencies, including annual modulation or at frequencies associated with solar rotation. Our analysis codes and datasets have been made publicly available.