Predicting Numerical Processing in Naturalistic Settings from Controlled Experimental Conditions

Machine learning research is interested in building models based on a training set that can then be applied to new data, whether this unseen data comes from new examples (e.g. New subjects, other tasks) or new features (e.g. Different modalities). In this work, we present a simple approach to transfer learning using intracranial EEG (also known as electrocorticographic, ECoG) data from three patients. More specifically, we aimed at detecting numerical processing during naturalistic settings based on a model trained with controlled experimental conditions. Our results showed significant prediction accuracy of numerical events in naturalistic settings when considering a priori knowledge of the target task.