Attractor dynamics underlie sequence working memory in data-constrained recurrent neural networks
Sequences of stimuli can be maintained in working memory by encoding each item, together with its position, as a distinct pattern of neural population acti