I think this is a really interesting area of machine learning. Some efforts have been made in ideas that are tangential to this one. Lots of papers in neuroevol

I think this is a really interesting area of machine learning. Some efforts have been made in ideas that are tangential to this one. Lots of papers in neuroevolution deal with evolving topologies. NEAT is probably the prime example http://nn.cs.utexas.edu/downloads/papers/stanley.ec02.pdf http://nn.cs.utexas.edu/downloads/papers/stanley.ec02.pdf and another paper I read recently called pathnet that is different but very interesting https://arxiv.org/abs/1701.08734 https://arxiv.org/abs/1701.08734.