In traditional machine learning, the goal is to “predict” the output of a function on previously unseen input data. The input data need not necessarily have a temporal dimension and the term “prediction” refers to the output of the learned function on a new data point. On the other hand, the task of forecasting is to predict the temporally future output of some function or the occurrence of an event. Time is thus a crucial component for forecasting. What makes forecasting particularly challenging is the fact that, from the (current) timepoint where a forecast is produced until the (future) timepoint for which we try to make a forecast, no data is available.
For this thesis, the candidate will have the opportunity to work with and/or develop a wide range of techniques (symbolic, statistical, machine learning) that can perform efficient event forecasting in real-time. Particular focus will be placed on combining deep learning and symbolic methods in order to develop a neuro-symbolic system for forecasting with the capacity to provide as its output forecasts that are not only accurate but also explainable and trustworthy. The developed system may be tested on a variety of application domains, from the analysis of biomedical data (epidemiological data, digital patients, evolution of patient trajectories) to the processing of moving objects trajectories and to the guidance of robotic swarms.
Bibliography:
Deliverable (besides the thesis itself): Publication quality report, 12 double-column pages