Training RNNs to do two-interval tasks with 2-Context or 2-Stimulus task structures. See Encoding time in neural dynamic regimes with distinct computational tradeoffs, 2022, Shanglin Zhou et al.
Folder 'model for tensorflow' contains codes for training the tasks in tensorflow (tested in version 1.15.0). Folder 'analysis' contains codes for the main meaures used in this paper: ASI/SSI at single unit level and SSI at population level. Prototypical neural dynamics are used to illustrate the implementation of those measures.