After following this presentation, I took 3 minutes to run stacknet and it moved me up about 25 positions on the leaderboard. By doing feature engeneering & selection I think performance will increase even more.
Very informative lecture! Thanks! Just one correction - AUC is actually not a measure of how accurate the predicted probabilities are, but of the ranking power of the model. If we transform every probability via x -> x^1000, we get all tiny probabilities, but the exact same AUC
After following this presentation, I took 3 minutes to run stacknet and it moved me up about 25 positions on the leaderboard.
By doing feature engeneering & selection I think performance will increase even more.
17:00 just to save you some time
Hey Team, Unless the speaker gives us access to slides we sadly don't have them. If we have them we do try and share them as the video goes live.
Very informative lecture! Thanks! Just one correction - AUC is actually not a measure of how accurate the predicted probabilities are, but of the ranking power of the model. If we transform every probability via x -> x^1000, we get all tiny probabilities, but the exact same AUC
when training the meta model, do you use labels too?
Amazing lecture! It's helpful for people who want to implement stacking!
Awesome! Very useful and informative talk! Thanks for sharing!
Just a suggestion. Please share the slide links as well.
Awesome！ learning a lot about the model stacking technique
Can we please have access to PPT slides ?
perfect and nice explanation !
Amazing, somehow it works!