My primary interest is RL, because I need it for my project (I don't think, that standard NN would work, because the enviroment changes so rapidly, that I can't use multiple agents, bec the results won't be comparable). I know some basics - I used TensorFlow in python for few times and I've built my own simple NN using matrix multiplication in unity few years ago. I just don't know much about RL and I don't know anything about neural networks in java, but I am trying to mod a java game and I didn't want to write the project in two languages before asking if there is some less wired solution.