This thesis presents novel machine learning algorithms that achieve state-of-the-art performance on a variety of electroencephalography (EEG) tasks, including decoding, classification, and unsupervised / semi-supervised artifact detection and correction. These algorithms are then used within the scope of an EEG experiment that explores how attention to multiple items modulates sensory representations. Using a signal detection paradigm, we demonstrate that attending to multiple items impacts the sensitivity of our participants, causing a sharp increase in false-alarm rates and only slightly decreasing hit-rate. We conclude that our behavioral and EEG decoding results contradict simultaneous attention guidance by multiple items (the multiple item template hypothesis).
Read
