Show HN: Never be 100% sure cuz that's overfitting 1 points by zipotm 5y ago ↗ HN A good relation to real life conclusions.
[–] minimaxir 5y ago ↗ That's not a good joke even from a statistical modeling perspective. Logistic models can absolutely make predictions with 100% confidence. [–] garmaine 5y ago ↗ Nothing has 100% confidence. We could be living in the matrix and it’s all a lie.Setting a probability equal to 0% or 100% is like dividing by zero.
[–] garmaine 5y ago ↗ Nothing has 100% confidence. We could be living in the matrix and it’s all a lie.Setting a probability equal to 0% or 100% is like dividing by zero.
[–] pizza 5y ago ↗ That's Cromwell's rule https://en.wikipedia.org/wiki/Cromwell%27s_ruleIf your prior is 0 then you can't do Bayes updates
3 comments
[ 2.9 ms ] story [ 15.9 ms ] threadSetting a probability equal to 0% or 100% is like dividing by zero.
If your prior is 0 then you can't do Bayes updates