Online SuperLearner

Jan 18, 2017

The goal of the Online SuperLearner (OSL) is to be able to make predictions and do causal inference on time series data in an (in machine learning terms) online fashion. In order to do so, it estimates a conditional density function for each of the covariates, the treatment variables, and the outcome. After these densities are fitted, OSL can simulate interventions and determine the outcome given such an intervention.

Although OSL itself is a generic implementation, the project currently focuses on applying it to medical data, such as time-series questionnaire data as collected from psychopathology research platforms, or physiological data as measured using sensors, or medical data collected in intensive care units. Projects that are of special interest are HowNutsAreTheDutch and ACTERRÉA.

The Online SuperLearer project is a collaboration between MAP5 (Université Paris Descartes), the Distributed Systems group (University of Groningen), and the developmental psychology group (University of Groningen),

-->