How to rerun a loop after checking accuracy ?

Comments

• sebastiaan

  May 2017 edited May 2017

  Hi,

  Say that you want to repeat block_loop when accuracy is less than 80%, then you could do the following:

  • Embed block_loop in another loop, here meta_loop
  • Set the repeat value for meta_loop very high (so that it's practically infinite)
  • Set the break-if statement for meta_loop to [acc] >= 80, and disable 'Evaluate on first cycle'

  Do you see the logic?

  

  Cheers,
  Sebastiaan

  
  
  Check out SigmundAI.eu for our OpenSesame AI assistant!
  
• aswv

  June 2017

  Works perfect.

  Thanks Sebastian...
• feffle

  August 2017

  Hi there,

  I am trying to do a similar thing and have tried this method but it seems to take the cumulative accuracy across all completed block repetitions. Is there a way to break the loop based only on accuracy of the most recent block?

  Thanks!
• eduard

  August 2017

  Hi,

  If you reset the feedback in the end (or the beginning) of a block, only the most recent block should be used for the computations. So just drag the reset_feeback item in the beginning of your block_sequence and you should be set.

  Good luck.

  Eduard

  
• feffle

  August 2017

  That worked a treat, thanks so much!
• Matanel Weissman

  December 2017 edited December 2017

  Hello, I have done what sebastiaan said and it worked, but I need an extra thing.

  I need please the experiment to end if the participant didn't reach to 80% accuracy. How can I do that?
  I need the exp to end after two tries to reach 80% acc - I know how to repeat if didn't get to 80% but I don't know how to end the exp.
• sebastiaan

  December 2017

  Hi @Matanel Weissman,

  I would use a simple inline_script that raises an Exception in this case. That will cause the experiment to exit. Like so:

  if var.acc < 80:
      raise Exception('Accuracy to low!')
  

  Cheers!
  Sebastiaan

  
  
  Check out SigmundAI.eu for our OpenSesame AI assistant!