Machine-translated by Google. Questions are authored in English.
Top

Discussion

What is the risk of training purely by self play?

  • A.Self play cannot generate enough games to train on
  • B.The program will always converge to random play
  • C.Self play requires more computation than human games
  • D.The program may exploit weaknesses only its own copies have

Answer: D

A pair of identical programs may settle into a narrow style, becoming excellent against each other and vulnerable to strategies neither explores. Randomised openings, population diversity and occasional external opponents mitigate it. Data volume is the method's strength, convergence to random play does not occur, and computation is the accepted cost. A training regime that defines its own test set risks optimising for a target that drifts away from the real one, which is a hazard shared with several other self supervised methods.

No comment is present. Be the first to comment.
Loading…

Post your comment

Please log in or register to post a comment.