Pages: 1 2 3 [4]
  Print  
Author Topic: Using Neural Networks in Spelunky?  (Read 5694 times)
KarjamP
Mossy
*
Posts: 3683


Meet Miijam!


View Profile WWW Email
« Reply #45 on: September 14, 2012, 11:44:19 AM »



Neural Networks are a method in implementing an ai. It can't learn on its own.

You'd have to use an algorithm like an generic algorithm to teach it something.
Logged

Meet Bodyless!

3DS friendcode: 4210 - 4681 - 1626



Join Dropbox!

BTC: 1M7MKFdwwrDqqowMUkrCXKhcEi7a16pwWy
Nathan
Mossy
*
Posts: 5087



View Profile Email
« Reply #46 on: September 14, 2012, 03:52:02 PM »

Whatever :/. Still A* with a learning algorithm would be best of all!
Logged

Wonkyth
Big Bossy
Mossy
*
Posts: 2126



View Profile
« Reply #47 on: September 15, 2012, 07:30:02 AM »

A* is a basic pathfinding algorithm, not so much AI.
Sure, it can be used within an AI system, but isn't much good on it's own.
Logged

Can you read the wry? Can you FEEL the wry?!
Nathan
Mossy
*
Posts: 5087



View Profile Email
« Reply #48 on: September 15, 2012, 01:31:28 PM »

I've seen many amazing Mario ai's made exclusively with A*
Logged

KarjamP
Mossy
*
Posts: 3683


Meet Miijam!


View Profile WWW Email
« Reply #49 on: September 16, 2012, 09:45:43 AM »

I've seen many amazing Mario ai's made exclusively with A*

It's an pathfinding algorithm It's useless on its own.

Besides, it's difficult to implement in my opinion (ESPECIALLY in a platformer). Tongue
Logged

Meet Bodyless!

3DS friendcode: 4210 - 4681 - 1626



Join Dropbox!

BTC: 1M7MKFdwwrDqqowMUkrCXKhcEi7a16pwWy
Nathan
Mossy
*
Posts: 5087



View Profile Email
« Reply #50 on: September 16, 2012, 12:01:23 PM »

I think using a Nural Network trained by a generic algorithm after a learning algorithm has watched you Play for a while would be the easiest way to go
Logged

KarjamP
Mossy
*
Posts: 3683


Meet Miijam!


View Profile WWW Email
« Reply #51 on: September 16, 2012, 11:16:20 PM »

^That doesn't make sense to me due to the way I understand how genetic algorithms works.

Genetic algorithms work by first doing random input for a certain amount of terms, evaluate which ones are the best, use the best ones as a basis for the next generation and maybe mutate them a bit just to make them more interesting.
Logged

Meet Bodyless!

3DS friendcode: 4210 - 4681 - 1626



Join Dropbox!

BTC: 1M7MKFdwwrDqqowMUkrCXKhcEi7a16pwWy
Nathan
Mossy
*
Posts: 5087



View Profile Email
« Reply #52 on: September 16, 2012, 11:20:22 PM »

It is entirely possible to put certain rules under how "random" a genetic algorithms base is
Logged

Pages: 1 2 3 [4]
  Print  
 
Jump to: