I learned a lot from this video, thanks 'mate'. I have 2 questions. 1. Several people have asked for the full source code, I believe you answered you still have it somewhere. Could you please make it available? If this is not possible, please let us know so we can stop begging. 2. I think you do agree that this is not a 2000 ELO chess engine. What you mean to say is that it was trained with >2000 ELO. Would your code reach 2000 ELO with better training? Or do you have ideas how the code could be improved?
Hello! This is a great video. In the ChessNet class, could you explain what PolicyNet is? And also where the predict(x) function comes from in choose_move()? Thanks so much!
@@moranreznik Sure. At 16:07 the very last line of the code block says move = predict(x), but the predict function was never defined in any of the other code shown. Could you explain the predict function, or do you have code showing it? Thanks
Hello! I loved your video. Would you please share your code with me so I can take a look at it? It would be of great inspiration for a little project I am trying to put together. Thanks!
@@moranreznik Hi, very good tutorial on the essential parts but I can't seem to be able to put together the missing pieces. If you could share your notebook, that would be very helpful !
Hi, i have watched your video so many times. Your project is very suitable for me but I am a beginner so there are many places that I do not understand and write myself. Can you give me the full code or link colab so I can read it?
Can you 1) link your code 2) show how you would do this using reinforcement learning. Currently I’m a network engineer and I try to get my hands on ML coding. To motivate myself I looking for AI useful that are actually add fun in your daily life. Your example was really great and motivated to kind of generate a improved version. Thank you again !
About using RL - hd tight, I will be releasing a series about deep RL soon. That being said, using it for a complex game like chess require A LOT of resources and time that I simply don't have
@@moranreznik thank you for the fast reply. Last question: what does form has t he output of your network ? Like does it output the typical chess syntax, e.g. b6 c5 ? Btw I found this really interesting because I play online chess with my cousin once a while and he always end up winning so I wanted to test the ML out, will do the implementation Sunday. Again thanks for the inspiration 😊
Not posting the source code + looking at the game leads me to believe that you want to take credit for making a 2000 ELO chess NN without having to actually make it. The problem is that that is HARD. If that is indeed the case then that is a shame because your method of encoding the chess board is interesting and so is how you encoded the moves with two 1-hot vectors. I learned from that, trying proceed from my value NN to something more robust that handles moves instead of just evaluating boards. You didn't need to brag for clicks and ego. You were plenty interesting without it. Or I am wrong. That is also a small possibility.
Let us just say that if a player with a 2000 elo played like this then i would urge him to call an ambulance. Something very very bad must have happened to his brain.
very nice tutorial! is there any chance you could link the full code?
Brilliant! Thank you. Requesting a residuals connection video
Hello, would you be able to share the colab notebook for this project? I would like to create a chess neural network of my own and this looks amazing!
@moranreznik could you please provide the full code?
Wow. I appreciate your preciseness and brevity, you done a great job, and even bigger effort to describe it so simply!
I learned a lot from this video, thanks 'mate'. I have 2 questions.
1. Several people have asked for the full source code, I believe you answered you still have it somewhere. Could you please make it available? If this is not possible, please let us know so we can stop begging.
2. I think you do agree that this is not a 2000 ELO chess engine. What you mean to say is that it was trained with >2000 ELO. Would your code reach 2000 ELO with better training? Or do you have ideas how the code could be improved?
Great tutorial! Is there any chance you could link the full code?
Xould you make a tutorial on how to make it uci compatible so we can play against it in chess intefaces and make it play against other bots
Great video. Really helpful!!!
can we get source code
Great Tutorial, Thanks!
Can u send us the full code link?
I tried replicating it but it did not end well. Can you please provide the full code :)
Hello! This is a great video. In the ChessNet class, could you explain what PolicyNet is? And also where the predict(x) function comes from in choose_move()? Thanks so much!
please provide exact time-stamps for your questions :)
@@moranreznik Sure. At 16:07 the very last line of the code block says move = predict(x), but the predict function was never defined in any of the other code shown. Could you explain the predict function, or do you have code showing it? Thanks
@@kieranodell4493 its a simple function, if I remember right, that excepts the board, inputs it into a NN and output a move, thats all.
@@moranreznik can you provide source code
17:01 losting Nc4 is not a 2000 ELO move.
Hello! I loved your video. Would you please share your code with me so I can take a look at it? It would be of great inspiration for a little project I am trying to put together. Thanks!
It sits in a colab notebook somewhere, ill try to find it
@@moranreznik thank you! That would be amazing
@@moranreznik Hi, very good tutorial on the essential parts but I can't seem to be able to put together the missing pieces. If you could share your notebook, that would be very helpful !
@@moranreznik have you found it yet? i also might need to use it to understand everything better
@@moranreznik did you find it?
Link to source code please
Hi, i have watched your video so many times. Your project is very suitable for me but I am a beginner so there are many places that I do not understand and write myself. Can you give me the full code or link colab so I can read it?
Excellent work
Can you 1) link your code 2) show how you would do this using reinforcement learning.
Currently I’m a network engineer and I try to get my hands on ML coding. To motivate myself I looking for AI useful that are actually add fun in your daily life. Your example was really great and motivated to kind of generate a improved version. Thank you again !
Thank you! :)
About using RL - hd tight, I will be releasing a series about deep RL soon. That being said, using it for a complex game like chess require A LOT of resources and time that I simply don't have
@@moranreznik thank you for the fast reply. Last question: what does form has t he output of your network ? Like does it output the typical chess syntax, e.g. b6 c5 ? Btw I found this really interesting because I play online chess with my cousin once a while and he always end up winning so I wanted to test the ML out, will do the implementation Sunday. Again thanks for the inspiration 😊
@@plemplem94 have you done it ?
Not posting the source code + looking at the game leads me to believe that you want to take credit for making a 2000 ELO chess NN without having to actually make it. The problem is that that is HARD. If that is indeed the case then that is a shame because your method of encoding the chess board is interesting and so is how you encoded the moves with two 1-hot vectors. I learned from that, trying proceed from my value NN to something more robust that handles moves instead of just evaluating boards. You didn't need to brag for clicks and ego. You were plenty interesting without it. Or I am wrong. That is also a small possibility.
Going through the video I agree with you more and more
W❤W! That’s Incredibly Awesome!
can we change the CNN by MLP, ig it can capture the whole board better then CNN
Why do you think so?
Is there a way to do this without python chess?
A chess ai without a chess package?
You can, but it will be much more difficult
@@moranreznik Yes,like can we create our own env, actions and learn and step functions.
does this perform better than minimax?
No.
Let us just say that if a player with a 2000 elo played like this then i would urge him to call an ambulance. Something very very bad must have happened to his brain.
the demo seemed more like 400 elo