search

memphis-iis / mofacts

8 stars 1 forks source link

Skips learning sessions #1341

Closed imrryr closed 1 year ago

imrryr commented 1 year ago

It goes right from the instructions to the survey in the next unit. Perhaps it thinks no items are available, e.g. if the clusterlist is not correctly loaded, bu still valid?

OptimTester3 preflight.zip

imrryr commented 1 year ago

the trace (available from) https://staging.optimallearning.org/experiment/mt3 seems to suggest that probability is never calculated perhaps because only cluster 0 is available (which isn't gonna work since that unit uses 3-40)

JRustyHaner commented 1 year ago

@imrryr while that was a problem, there's another: the currentDeliveryParameters.optimalThreshold was returning as 0

Math.log(0/1-0) gives -infinity, which is impossible to find distances for.

I don't know how you want to fix that, since I'm not sure you were aware of that possible bug

imrryr commented 1 year ago

well, the value is set in the tdf, so this other problem is a good catch, since it is potential, but I don't see how it happens here since it is set to .7

JRustyHaner commented 1 year ago

@imrryr I got it, the translation in currentTestingHelpers is running parseInt on it,

The integer value of 0.7 is 0, leading to Math.log(0/1-0) to -Inf

imrryr commented 1 year ago

After thinking about this, if an experimenter leaves this unset, I think it is best considered a bug in the tdf and it should fail. So please make sure there is no default and it will error if this is not provided either in Parameters[1] or optimalThreshold