Deep Neural Networks for Forecasting Single-Trial Event-Related Neural Activity

In this work, we propose a deep neural network to forecast single-trial event-related electroencephalographic (EEG) activity using observed pre-event EEG data. Forecasting eventrelated potentials (ERPs) have a number of potential benefits for brain-computer interface (BCI) systems. Accurate predictions of neural responses can reduce the latencies of neural signal classification algorithms and closed-loop feedback systems. Various models of ERPs propose that there exists a causal dependency between post-event neural responses and ongoing pre-event neural dynamics. Accordingly, we implement a deep neural network that extracts features from pre-event data in order to predict a single-trial ERP. To capture the variability of a single trial, the network constructs the post-event waveform in two parts: 1) generating ongoing neural activity and 2) generating event-related components comprising the ERP. We evaluate our model by forecasting 500 milliseconds of single channel postevent data from a Rapid Series Visual Presentation (RSVP) task. Our results indicate a significant increase in forecasting performance compared to baseline methods, suggesting that deep neural networks can extract informative features from pre-event EEG data in order to generate a prediction of the post-event waveform.