Conference on Cognitive Computational Neuroscience

Humans are constantly processing scene information from their environment to make meaningful decisions. However, it remains unknown which of the emerging neural representations are relevant for decision-making, and how we can model these representations and their link to behaviour in a unified way. Here, we approached these questions empirically and via computational modelling using deep neural networks. First, we correlated electroencephalography (EEG) data and categorization reaction times and observed a brain-behaviour correlation between ~100 ms and ~200 ms after stimulus onset. Second, to provide a model of scene categorization which integrates both neural and behavioural data, we evaluated a recurrent convolutional neural network (RCNN) and observed similarities between humans and the network in terms of neural representations, behaviour, and the brain-behaviour relationship. Altogether, our results provide an empirical and computational account of scene categorization in humans, benefiting both cognitive and computational neuroscience for future exploration and explanation of human decision-making.