In recent works, neural-nets implementations of reaction-diffusion type approaches for geometric flows have been proposed and successfully implemented. In this talk, we review some of these approaches in the simplest case of the Mean Curvature Flow of boundaries, and the more complex case of the MC flow of non-oriented boundaries, then, we discuss analytical variants with similar properties and behaviour. This is joint work with Elie Bretin, Simon Masnou (Lyon), Daniele De Gennaro (Milano), Massimiliano Morini (Parma).