I am working on trainable morphogenic functions for reaction-diffusion proceses. I.e., using Pytorch, the parameters of a neural net representing the reaction p

I am working on trainable morphogenic functions for reaction-diffusion proceses. I.e., using Pytorch, the parameters of a neural net representing the reaction part are optimized to so the diffusion process converges in a complex, predefined structure. This is only a pet project, but I am fascinated by this because I imagine that in the future this could be used to design synthetic biological structures. All that will be needed is a compiler that translate the function to DNA...