Should we also submit the result with Olga method?
", "messages": [ { "id": "182957", "postDate": "05/16/2017 13:53:31", "content": "Should we also submit the result with Olga method?
", "rawMarkdown": "Should we also submit the result with Olga method?", "votes": null }, { "id": "182972", "postDate": "05/16/2017 15:05:36", "content": "r4m0n mentioned in https://www.kaggle.com/c/intel-mobileodt-cervical-cancer-screening/discussion/32213 that he'll release the truth labels for stage 1 if organizers don't make things clear about LB mining. \nNo statement was made about that so I think it would be fair to release the labels as chaos on the LB is growing anyway and this way we would be able to avoid unfair advantage given to those who decided to mine.
", "rawMarkdown": "r4m0n mentioned in https://www.kaggle.com/c/intel-mobileodt-cervical-cancer-screening/discussion/32213 that he'll release the truth labels for stage 1 if organizers don't make things clear about LB mining. \nNo statement was made about that so I think it would be fair to release the labels as chaos on the LB is growing anyway and this way we would be able to avoid unfair advantage given to those who decided to mine.", "votes": null }, { "id": "182973", "postDate": "05/16/2017 15:11:59", "content": "They have replied about the Stage 1 data at least, though they don't seem to care about people violating the rules of the competition and posting data they didn't mine themselves to the LB.
\n\nAs they are going to release the stage 1 data themselves, I'll refrain from doing that myself.
", "rawMarkdown": "[They have replied about the Stage 1 data at least][1], though they don't seem to care about people violating the rules of the competition and posting data they didn't mine themselves to the LB.\n\nAs they are going to release the stage 1 data themselves, I'll refrain from doing that myself.\n\n\n [1]: https://www.kaggle.com/c/intel-mobileodt-cervical-cancer-screening/discussion/30205#179822", "votes": null }, { "id": "182985", "postDate": "05/16/2017 15:57:21", "content": "Ah, I have missed that one, sorry! The unfortunate thing is the fact that either you can mine the LB and get bigger set for training/validation or stick to original data and be at potential disadvantage.
", "rawMarkdown": "Ah, I have missed that one, sorry! The unfortunate thing is the fact that either you can mine the LB and get bigger set for training/validation or stick to original data and be at potential disadvantage.", "votes": null }, { "id": "187345", "postDate": "05/31/2017 02:11:01", "content": "Stupid question: what is \"leaderboard mining\"?\nHow do you think they managed to get a 0.00000 score?
", "rawMarkdown": "Stupid question: what is \"leaderboard mining\"?\nHow do you think they managed to get a 0.00000 score?", "votes": null }, { "id": "187392", "postDate": "05/31/2017 05:13:51", "content": "Leaderboard mining can be done like this.
", "rawMarkdown": "Leaderboard mining can be done like [this][1].\n\n\n [1]: https://www.kaggle.com/olegtrott/the-perfect-score-script", "votes": null } ], "comments": [ { "id": 182972, "author_name": "wrosinski", "author_url": "", "post_date": "05/16/2017 15:05:36", "content": "r4m0n mentioned in https://www.kaggle.com/c/intel-mobileodt-cervical-cancer-screening/discussion/32213 that he'll release the truth labels for stage 1 if organizers don't make things clear about LB mining. \nNo statement was made about that so I think it would be fair to release the labels as chaos on the LB is growing anyway and this way we would be able to avoid unfair advantage given to those who decided to mine.
", "votes": null, "replies": [ { "id": 182973, "author_name": "mumech", "author_url": "", "post_date": "05/16/2017 15:11:59", "content": "They have replied about the Stage 1 data at least, though they don't seem to care about people violating the rules of the competition and posting data they didn't mine themselves to the LB.
\n\nAs they are going to release the stage 1 data themselves, I'll refrain from doing that myself.
", "votes": null, "replies": [] }, { "id": 182985, "author_name": "wrosinski", "author_url": "", "post_date": "05/16/2017 15:57:21", "content": "Ah, I have missed that one, sorry! The unfortunate thing is the fact that either you can mine the LB and get bigger set for training/validation or stick to original data and be at potential disadvantage.
", "votes": null, "replies": [] } ] }, { "id": 187345, "author_name": "edoven", "author_url": "", "post_date": "05/31/2017 02:11:01", "content": "Stupid question: what is \"leaderboard mining\"?\nHow do you think they managed to get a 0.00000 score?
", "votes": null, "replies": [ { "id": 187392, "author_name": "oysteijo", "author_url": "", "post_date": "05/31/2017 05:13:51", "content": "Leaderboard mining can be done like this.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "182957": "Should we also submit the result with Olga method?", "182972": "r4m0n mentioned in https://www.kaggle.com/c/intel-mobileodt-cervical-cancer-screening/discussion/32213 that he'll release the truth labels for stage 1 if organizers don't make things clear about LB mining. \nNo statement was made about that so I think it would be fair to release the labels as chaos on the LB is growing anyway and this way we would be able to avoid unfair advantage given to those who decided to mine.", "182973": "[They have replied about the Stage 1 data at least][1], though they don't seem to care about people violating the rules of the competition and posting data they didn't mine themselves to the LB.\n\nAs they are going to release the stage 1 data themselves, I'll refrain from doing that myself.\n\n\n [1]: https://www.kaggle.com/c/intel-mobileodt-cervical-cancer-screening/discussion/30205#179822", "182985": "Ah, I have missed that one, sorry! The unfortunate thing is the fact that either you can mine the LB and get bigger set for training/validation or stick to original data and be at potential disadvantage.", "187345": "Stupid question: what is \"leaderboard mining\"?\nHow do you think they managed to get a 0.00000 score?", "187392": "Leaderboard mining can be done like [this][1].\n\n\n [1]: https://www.kaggle.com/olegtrott/the-perfect-score-script" }, "source": "meta" }