Use this topic to post any questions or request for clarification about the competition rules.
Do I have to use all the curves in the prediction? or it is ok to use a subset of the curves. My approach is not all of the logs respond to a lithology changes.
@dianaceroa You do not have to use all the curves. The model in the example notebook discards several curves. If you end up submitting for the final scoring, then your model must pick it’s relevant curves as part of the preprocessing step.
I have a question about the scoring on the private dataset. I do not fully understand how it will be done and see 2 options
- Top10 teams on Public Leaderboard are invited to submit their code/models so that organizers can run the code/models on the private dataset.
- Submission dataset already contains public and private dataset, so results on the private leaderboard will be available right after the competition ends (it is not necessary to run any code to get Private Leaderboard results). After that top10 teams on the Private Leaderboard will be invited to submit their code/models.
Please clarify which option is correct if any. Thanks in advance!
Hi @dzikr! The correct answer is option 1. More information about how to submit models for testing will be sent out to the top 15 participants at the end of the contest later this week.
While working with this data, a lot of assumptions have been made silently. For example, one of them is that number of wells in the training data is not that large (10 in current public leaderboard dataset). Can it be assumed that in private dataset there are about 10 wells too?
@dzikr yes about ten wells