I am looking at developing an algorithm that would automatically grow the structure of a neural network by adding/deleting units within a layer, or adding layers as necessary.
I researched the topic and only found literature from before 2000. I was somewhat aware of it, being old enough to have come across some of them when they were published. What I found is mostly interested in neuron-by-neuron construction. I didn't find anything layer-by-layer.
I have searched for more recent papers, including in the deep learning space and have drawn a blank.
I wrote a draft blog article laying out my thoughts (https://emmanuel-r8.github.io/post/neural-network-incremental-growth.html). I will carry on researching the topic. But I am looking for references that would tell me that it is either a blind alley or expand my horizons.
