The objective is to provide a software tool that analyzes non-invasive EEG signals and, at regular intervals, forecasts how likely a seizure will be in the nearest future. To achieve this goal, students must design a deep learning algorithm, e.g., by using convolutional neural networks, to extract EEG spectral features and/or functional connectivity features over time and estimate the seizure likelihood. The project will be conducted in MATLAB using scalp EEG data from a previous study.