Hacker News new | past | comments | ask | show | jobs | submit login

I have reasonable suspicion that it wouldn't be nearly as easy to train without "spinning" move which basically gives invulnerability against all enemies sans their bullets, iirc.



I see this videoclip as a great introduction to neural networks and genetic algorithms (something I didn't know existed in ML) rather than a simple "computer completes videogame level". You're probably right nonetheless (I can't remember the mechanics of this mario game) :)


Actually for Genetic Algorithms and/or Evolutionary computation it is not. :P


As far as I can tell, it would only require longer time (more iterations) for the network to learn, but it would still work.

The main problem I see is that the algorithm is learning from the same level (or set of levels). It would be interesting to see what happens if you give it some completely new level to play.


Wait, does it really give you invulnerability? In the video Mario never runs into a single enemy while spin jumping.


Not exactly. Some enemies that would kill you with a normal jump when you land on them, Mario will bounce off of with a spin jump.


A similar work was done for Super Mario on NES (along with other NES games): https://www.youtube.com/watch?v=xOCurBYI_gY




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: