Over the last years machine learning has caught on big-time. Although many of the statistical methods have been well established for decades, their applications have seen a real revolution. This is mostly due to access to the vast amounts of data collected through web-services and social media as well as substantial increases in computing power. Machine learning is primarily powerful for all kind of prediction tasks. Nonetheless, it is also likely to lead to new directions in epidemiological research , even where the focus is more on disentangling causality. Further, familiarity with basic machine learning gives invaluable new perspectives on statistical analysis more generally, and should be a part of the toolbox of anyone doing modern data analysis.
In this talk we will address some topics within the field of machine learning. Christian Madsen will give a brief overview of machine learning and epidemiology. Kåre Bævre will talk about prediction/classification with examples from his work on using deep neural nets for reading of handwritten census data. Yunsung Lee will talk about his work on using machine learning methods on DNA-methylation data.