Found this neat NEAT library, unfortunately it seems to severely lack documentation or up-to-date tutorials with clear explanations.
I've been thinking about trying out NEAT to potentially create some simple image / sound compression algorithm, or maybe to reconstruct images and sound.
I think regular NEAT might be just good enough since I could just process images in small 16x16 or 32x32 pixel blocks or chunks (HyperNEAT would probably be better though but is more complicated).
I feel like it should be possible to use NEAT to create some sort of image compression or upscaling algorithm.
There is however, a severe lack of information on how to do something simple such as using NEAT to generate or transform small 8x8 or 16x16 images.
I saw something about NEAT being used to generate random abstract images where fitness was determined by which images the user liked the most but no example code or documentation on what does what.
This usage of NEAT however requires that the fitness evaluator waits for user input and cannot proceed without which requires a different approach to how it's trained.
Secondly, I was interested in potentially using NEAT in a game engine such as Godot (smarter or adaptive enemy AI), but with a lack of documentation I do not know how to integrate NEAT into the update / game loop of a game engine such as Godot.
Using SharpNEAT in a game requires a lot of waiting on the physics engine or for actors to walk towards a goal which means SharpNEAT needs to adapt to a system where things take time or cannot happen instantly.
I have quite a bit of programming experience but have always found it notoriously difficult to integrate a library into an existing project without completely redesigning or rewriting things from scratch.
Found this neat NEAT library, unfortunately it seems to severely lack documentation or up-to-date tutorials with clear explanations.
I've been thinking about trying out NEAT to potentially create some simple image / sound compression algorithm, or maybe to reconstruct images and sound.
I think regular NEAT might be just good enough since I could just process images in small 16x16 or 32x32 pixel blocks or chunks (HyperNEAT would probably be better though but is more complicated).
I feel like it should be possible to use NEAT to create some sort of image compression or upscaling algorithm.
There is however, a severe lack of information on how to do something simple such as using NEAT to generate or transform small 8x8 or 16x16 images.
I saw something about NEAT being used to generate random abstract images where fitness was determined by which images the user liked the most but no example code or documentation on what does what.
This usage of NEAT however requires that the fitness evaluator waits for user input and cannot proceed without which requires a different approach to how it's trained.
Secondly, I was interested in potentially using NEAT in a game engine such as Godot (smarter or adaptive enemy AI), but with a lack of documentation I do not know how to integrate NEAT into the update / game loop of a game engine such as Godot.
Using SharpNEAT in a game requires a lot of waiting on the physics engine or for actors to walk towards a goal which means SharpNEAT needs to adapt to a system where things take time or cannot happen instantly.
I have quite a bit of programming experience but have always found it notoriously difficult to integrate a library into an existing project without completely redesigning or rewriting things from scratch.