I am developping a neural-network based Chess engine

Are there any plans to extend this work to the other variants played here? It would be nice to see what they can do!

given the amout of computing resources needed it is very unlikely that anyone would be interested on variants. In particular you need people contributing they CPU/GPU for training. Hard to see that happen for variants

Apart from Chess360 they are all boring and from ground up trained network should do fine on that

hello, I know how I first started to play chess with no prior knowledge, which got me beating 1820s in first few games. What I did could be replicated by a computer in a way very easily. My process of move development was to first limit the use of my opponents knights then to go on to bishops. You can easily stifle an opponents options doing this and I looked at every single move for about 4 moves, not understanding any concepts like I do now.

looking for pawnbreaks is another 'process'