Es fácil, lo primero que debes hacer es mejorar el oro. Luego defiendete con soldados de 50 si el saca los de 50 o los de 100. Si saca los de 150 defiende con uno de 100. Y ahorra para mejorar el oro al maximo y luego gana a puros de 200. Alto juegito
Hey mas que interesante tu proyecto! Muy buena idea usar la red neuronal :D Si queres pasa y prueba pixel commando, mi proyecto en curso, seria genial! nos vemos.
This is pretty good. I dont think that the neural networks for different difficulties were different enough. Easy took 2 tries, Normal took 2 tries, Hard took 2 tries. But this is fun to play anyway!
Thank you very much! Yes... the truth is that the difficulty is not very noticeable, each neural network has learned some tactics but they are easy to counteract
because its a nueral network the more i play the more it will learn right? and it should get harder for me right? Will it also get harder for other people?
well, technically it could be done, but it complicates the development quite a bit, so in this prototype, the network doesn't learn no matter how much you play
so that's the idea, you can't predict 100% the behavior of a neural network, normal AI is scripted by the programmer, while a neural network acquires its behavior through training, I trained it for about 15 hours
Maybe let us train it in the game itself. And even set learning parameters such as the size of the neural network. The learning rate. Or even the amount of iterations simultaneously learning. But just letting us train it would be great.
← Return to game
Comments
Log in with itch.io to leave a comment.
first try every difficulty
me dice que un error ocurrio y no puedo jugarlo
ta bueno
esta faci
Sympa
Es fácil, lo primero que debes hacer es mejorar el oro. Luego defiendete con soldados de 50 si el saca los de 50 o los de 100. Si saca los de 150 defiende con uno de 100. Y ahorra para mejorar el oro al maximo y luego gana a puros de 200. Alto juegito
Es una buena idea tu juego, me gusto, me entretuvo por un rato, espero que tengas mas suscriptores ya que tu canal es muy bueno
Hey mas que interesante tu proyecto! Muy buena idea usar la red neuronal :D Si queres pasa y prueba pixel commando, mi proyecto en curso, seria genial! nos vemos.
Pues, al igual que la red neuronal me tomo 2 intentos ganarle a modo dificil xd
me encanto tu juego
major balancing issues on the unit it seems its way too easy if you get extra gold
easy
The 100 cost unit seems a little over powered imo. It takes down the tower so fast.
I keep playing hard mode over and over, and sometimes its 2 minutes and sometimes its 6 (I always win though)
This is pretty good. I dont think that the neural networks for different difficulties were different enough. Easy took 2 tries, Normal took 2 tries, Hard took 2 tries. But this is fun to play anyway!
Thank you very much! Yes... the truth is that the difficulty is not very noticeable, each neural network has learned some tactics but they are easy to counteract
because its a nueral network the more i play the more it will learn right? and it should get harder for me right? Will it also get harder for other people?
well, technically it could be done, but it complicates the development quite a bit, so in this prototype, the network doesn't learn no matter how much you play
so why does it react differently in the exact same situation?
when i start i always do the same thing but it sometimes does different things. Isnt that just a normal AI?
so that's the idea, you can't predict 100% the behavior of a neural network, normal AI is scripted by the programmer, while a neural network acquires its behavior through training, I trained it for about 15 hours
Maybe let us train it in the game itself. And even set learning parameters such as the size of the neural network. The learning rate. Or even the amount of iterations simultaneously learning. But just letting us train it would be great.