Statistical learning and advanced computational methods are increasingly being applied to problems in atmospheric chemistry. These methodological advances combined with increasing satellite and in-situ observations provide new opportunities to improve model predictions and our understanding of atmospheric processes. This session aims to provide a forum for research on data-driven techniques for understanding the sources and concentrations of air pollutants (NOx, SOx, CO, O3, PM2.5, etc.) and greenhouse gasses (CO2, CH4, N2O, etc.). We encourage submissions on both the application of data-driven methods for interpreting new atmospheric composition datasets and the development of new methods in machine learning, statistical inference, data assimilation, and cloud computing.