Slow manifolds in recurrent networks encode working memory efficiently and robustly

Open in new window