12 Replies - 598 Views - Last Post: 24 April 2017 - 01:56 PM

#1 mdj  Icon User is offline

  • New D.I.C Head

Reputation: 0
  • View blog
  • Posts: 6
  • Joined: 24-April 17

Neural Network and qlearning. Evolution

Posted 24 April 2017 - 10:40 AM

Posted Image


My first experience with neural network and q-learning.
In next step I want to add genetic algorithm and expand world, and sure to add enemies.

Please watch the video:

https://www.youtube.com/watch?v=6HILhAcwXLE
Is This A Good Question/Topic? 0
  • +

Replies To: Neural Network and qlearning. Evolution

#2 modi123_1  Icon User is online

  • Suitor #2
  • member icon



Reputation: 13382
  • View blog
  • Posts: 53,406
  • Joined: 12-June 08

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 10:48 AM

Please, by all means, explain your project, what you are learning, what's going on in the video, etc.

Certainly helps not make this one post just a scatter shot for spam. ;) :^:

So far also posted here and here.
Was This Post Helpful? 1
  • +
  • -

#3 mdj  Icon User is offline

  • New D.I.C Head

Reputation: 0
  • View blog
  • Posts: 6
  • Joined: 24-April 17

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 10:53 AM

View Postmodi123_1, on 24 April 2017 - 10:48 AM, said:

Please, by all means, explain your project, what you are learning, what's going on in the video, etc.

Certainly helps not make this one post just a scatter shot for spam. ;)/> :^:/>


I did try to find the button to edit the topic, but .. where is it?

Ok.
Agents, that you can see on the field have brain (neural network), witch learning to eat (red points on the field). If agent's brain not enough "smart" he can't eat enough and he finally die, because food influence on their health.
As you can see on beginning of the video agents not understood what's going on too. And after a while they start to try eat.

In other words it is demonstration of Artificial Intelligence based on neural network and q-learning.
Was This Post Helpful? 0
  • +
  • -

#4 modi123_1  Icon User is online

  • Suitor #2
  • member icon



Reputation: 13382
  • View blog
  • Posts: 53,406
  • Joined: 12-June 08

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 11:03 AM

You haven't been on the site long enough to have the edit option available.

What, exactly, is this model it is trying to learn/update/reinforce? Seems like some anthropomorphize of the agents is going on.
Was This Post Helpful? 0
  • +
  • -

#5 mdj  Icon User is offline

  • New D.I.C Head

Reputation: 0
  • View blog
  • Posts: 6
  • Joined: 24-April 17

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 11:07 AM

If it badly just delete the topic and I'm sorry.
Was This Post Helpful? 0
  • +
  • -

#6 modi123_1  Icon User is online

  • Suitor #2
  • member icon



Reputation: 13382
  • View blog
  • Posts: 53,406
  • Joined: 12-June 08

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 11:13 AM

Brah, I am just asking questions about it. This started as a drive by post this, but I am actually interested in what is going on behind the scenes.

Is the code out somewhere, like github, where I can poke around at it?
Was This Post Helpful? 0
  • +
  • -

#7 mdj  Icon User is offline

  • New D.I.C Head

Reputation: 0
  • View blog
  • Posts: 6
  • Joined: 24-April 17

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 11:17 AM

View Postmodi123_1, on 24 April 2017 - 11:03 AM, said:

What, exactly, is this model it is trying to learn/update/reinforce? Seems like some anthropomorphize of the agents is going on.


Agents will "smarter" if process will take a long time, but the program haven't genetic algorithm and saving weights of the neural network yet. Actually I want to find people who also do this for experience exchange.
Was This Post Helpful? 0
  • +
  • -

#8 modi123_1  Icon User is online

  • Suitor #2
  • member icon



Reputation: 13382
  • View blog
  • Posts: 53,406
  • Joined: 12-June 08

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 11:20 AM

You really don't need to quote the post.. just use the reply button. :^:

Okay.. so what does this 'smarter' mean? How is that expressed? It doesn't seem like the agents are any more adept at finding food.. I mean sure they randomly hit something and spin in a circle. That doesn't seem like "learned" behavior.
Was This Post Helpful? 0
  • +
  • -

#9 mdj  Icon User is offline

  • New D.I.C Head

Reputation: 0
  • View blog
  • Posts: 6
  • Joined: 24-April 17

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 11:25 AM

View Postmodi123_1, on 24 April 2017 - 11:13 AM, said:

Brah, I am just asking questions about it. This started as a drive by post this, but I am actually interested in what is going on behind the scenes.

Is the code out somewhere, like github, where I can poke around at it?


The code there is on github, but it's older version without neural network:
https://github.com/U...MDJ/EatingWorld

At the end of this project agents will learn to survive in this world. They need to eat and run from the enemies
Was This Post Helpful? 0
  • +
  • -

#10 modi123_1  Icon User is online

  • Suitor #2
  • member icon



Reputation: 13382
  • View blog
  • Posts: 53,406
  • Joined: 12-June 08

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 11:32 AM

Where in that code is the part where the learning/reinforcement that food is a positive thing and/or more is better?
Was This Post Helpful? 0
  • +
  • -

#11 modi123_1  Icon User is online

  • Suitor #2
  • member icon



Reputation: 13382
  • View blog
  • Posts: 53,406
  • Joined: 12-June 08

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 12:28 PM

Looks like Lenny broke another one George. :/
Was This Post Helpful? 0
  • +
  • -

#12 mdj  Icon User is offline

  • New D.I.C Head

Reputation: 0
  • View blog
  • Posts: 6
  • Joined: 24-April 17

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 01:55 PM

I said that it's older version and there isn't neural network, I can explain you how it works, but can't show the code.
There are q functions which are neural network results. These q functions defines action which will do agent. Q-f. changes on the assumption of signals. If neural network's result was good for agent (he just ate or was near food) neural network is encouraged and in next time with same signals agent will try to do same set of actions for food production, or is punished if it crashes into the wall or in future will meet the enemy. I can give you some math functions and explanations which I used to use in the program if you interested or better you will break your self?
Was This Post Helpful? 0
  • +
  • -

#13 modi123_1  Icon User is online

  • Suitor #2
  • member icon



Reputation: 13382
  • View blog
  • Posts: 53,406
  • Joined: 12-June 08

Re: Neural Network and qlearning. Evolution

Posted 24 April 2017 - 01:56 PM

Sure.. explain more and dump the math and project flow out here.

I would totally look it over.
Was This Post Helpful? 0
  • +
  • -

Page 1 of 1