This video referred to one aspect many of the videos about the same subject do not: You get one model per fold (one set of "fitted" parameters, one "RMSE" if that is what you are using to evaluate that model, one set of predicted features, etc), not one final model so, as the author said, you use it to evaluate how a certain type of model can perform by averaging the total of models (averaging the statistic you are using to evaluate how good is the model) you get when doing one model training per fold, NOT to find the final model parameters. It is almost never mentioned this tiny detail. And I see many people, like myself, wondering what the end result of this method is and its usage.
Impressed by the quality and enjoy how concise and short your videos are!
Im glad I found this. Easy to understand. Thank you
the best and easy explanation ever, good job
Thank you thank you.. I was puzzled with how do we decide which model to adopt until I saw this video.
Perfectly explained, thanks!
This video referred to one aspect many of the videos about the same subject do not: You get one model per fold (one set of "fitted" parameters, one "RMSE" if that is what you are using to evaluate that model, one set of predicted features, etc), not one final model so, as the author said, you use it to evaluate how a certain type of model can perform by averaging the total of models (averaging the statistic you are using to evaluate how good is the model) you get when doing one model training per fold, NOT to find the final model parameters. It is almost never mentioned this tiny detail. And I see many people, like myself, wondering what the end result of this method is and its usage.