I am wondering if I could get some clarification on Part 1 (a) and Part 1 (c) of the problem set.
As for Part 1 (a), I can't seem to find the cutoff point/threshold where the submission would not be judged. Please correct me if I'm wrong. My impression from reading the Netflix rule is that all submissions would be judged, but for a submission to be qualified for winning, it has to achieve 10% improvement of RMSE. But, I couldn't quite figure out the cutoff point for poor performance.
As for Part (c), could you please elaborate more on this part of the question: "What characteristic of one model relative to other models made it improve the overall prediction when blended with the other models?"
Do you mean that we should describe a model in the ensembles?
I am wondering if I could get some clarification on Part 1 (a) and Part 1 (c) of the problem set.
As for Part 1 (a), I can't seem to find the cutoff point/threshold where the submission would not be judged. Please correct me if I'm wrong. My impression from reading the Netflix rule is that all submissions would be judged, but for a submission to be qualified for winning, it has to achieve 10% improvement of RMSE. But, I couldn't quite figure out the cutoff point for poor performance.
As for Part (c), could you please elaborate more on this part of the question: "What characteristic of one model relative to other models made it improve the overall prediction when blended with the other models?"
Do you mean that we should describe a model in the ensembles?
Any hints would be much appreciated!