There are forums out there on programming neural network chess engines. There a several non commercial NN engines out there, with more being programmed all the time. It's probably still a small community at the present. Certainly some of the people on the forums worked on Google's Alpha Chess Zero engine.
-They don't give out code, but they do give out enough information that you can understand the underlying concepts and how to figure out the coding yourself. So far, I think Komodo and Fritz have NN engines on a commercial level. The Fritz one in particular is a true NN. The Komodo is sort of a hyrid, in that they program a certain level of knowledge into it, so that it has a leg up. The way Fritz works, the technology isn't there yet to really make a single PC version of NN engine that plays on the level of Stockfish or Komodo. But the Fritz NN version is just a bonus to the regular Fritz engine if you buy Fritz.
Nobody has come close to Google's Alpha Chess Zero, but that's ok, neural network engines are in it's preschool years, and nobody outside of Google knows what hardware it was running on, which most people expect was phenomenal, both in terms of raw processing power, and the amount of ram and storage capacity they have available.
In any event, all neural network engines need an opening book they're forced to use, or they will play a very limited opening repertoire.
So I started brainstorming with the idea of making my own personal chess bot/engine. I am looking to make the bot a neural network based engine. This is a coding project I put on myself to kill some time and would love to see my own creation work it's way up ratings. I'm not looking for this bot to be up to par with AlphaZero or Leela Chess Zero. This is to be a simple bot that can most likely reach, ~1700? I think with that I would be happy.
However, I lack the knowledge of the machine learning path. I know c++, Java and a little bit of python. However, I know this isn't enough. I have a friend of mine give me some advice from a cs(computer science) point of view, but I would like to know from a chess community/chess point of view what I should look for. What resources or sources can I start with to begin my journey. I know the task isn't gonna be easy nor is it gonna be short. I don't mind it taking some time although ideally I would like to have it done by August if I were to start now, but I know I'll be busy with other stuff that I might not finish at my ideal date especially if I have to learn a lot/do tons of research and reading.
Potential questions to answer:
1. Is it worth it to make a NN engine or should I hard code it?
2. What are some baby steps to take to approach the final result? Is there something I should do first before making a chess bot? Is there a certain road to follow?
3. Any libraries, pages, sources, or videos you recommend?
4. Is it worth the time and investment?
5. Do I need a strong computer? (I have a powerful one, but if it's true, how strong does it need to be? ) I've been told ML requires plenty, and I mean PLENTY of data to be efficiently good and reliable.
6. Can this be achieved self taught/alone or is extentive professional knowledge/education required?
That pretty much sums up my various questions and I'm sure there's more. If you made it this far I do ask please not to flame or be negative. I lack sufficient information to be well informed about cs and chess, but thought this would be a great and fun project for myself. If there is something wrong or any information that is incorrect please just let me know. I am pretty much starting this journey from scratch and want to see how far I can go. I am willing to put in the time and effort into learning whatever it is I need to learn/do. I know with enough dedication and effort I can make it happen even if I'm not that advanced. Thank you for taking the time for reading this at least and if you reply I thank you in advance. Much appreciated, stay well, stay safe, and stay blessed everyone.