主题：【原创】围绕脑科学而发生的若干玄想 -- 鸿乾

The authors have missed out some of the developments in the recent decade. There's, for example, Schmidhuber's compressibility gradient, Ay and Der's homeokinesis predictive information maximization, Klyubin et al.'s empowerment method, and the underlying Touchette/Lloyd formalism for describing entropic bounds on open and closed-loop control and many more. Plus, the authors do not seem to mention in their paper Linsker's criticism of Dewar's MEPP derivation. In isolation, it is difficult to put in perspective how the approach sits in the context of the existing work and how much it depends on postulated MEPP-type assumptions. Which are perfectly fine to make, but it'd be nice to see how it fits into other closely related models which are already studied in some detail.