There may be ways to improve prediction quality by training our models using data beyond the given training set. First choice, naturally, may be to search the sites of the organizations mentioned in the competition overview like USDA, NSF, CHORUS, NIH etc. - thanks to the organizers for this hint. But there surely are many more sources out there. Besides, external data is indirectly used by those of us who import pre-trained models.
\nIt may help to establish guidelines on use and disclosure in this respect to avoid surprises towards the end of the competition. Could @julialane or someone else in the organizers team clarify the external data use aspects.
There may be ways to improve prediction quality by training our models using data beyond the given training set. First choice, naturally, may be to search the sites of the organizations mentioned in the competition overview like USDA, NSF, CHORUS, NIH etc. - thanks to the organizers for this hint. But there surely are many more sources out there. Besides, external data is indirectly used by those of us who import pre-trained models.
\nIt may help to establish guidelines on use and disclosure in this respect to avoid surprises towards the end of the competition. Could @julialane or someone else in the organizers team clarify the external data use aspects.
Hi @poedator - thanks for your questions. We're discussing with the host and we'll post our decision as soon as it's made.
", "rawMarkdown": "Hi @poedator - thanks for your questions. We're discussing with the host and we'll post our decision as soon as it's made.", "votes": null }, { "id": "1272660", "postDate": "04/13/2021 17:04:34", "content": "We created a post to capture external data sources. External data is allowed but it must be freely and publicly available and must be sited before the competition deadline.
", "rawMarkdown": "We created a post to capture[ external data sources](https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/232414). External data is allowed but it must be freely and publicly available and must be sited before the competition deadline.", "votes": null } ], "comments": [ { "id": 1271670, "author_name": "philculliton", "author_url": "", "post_date": "04/12/2021 19:15:05", "content": "Hi @poedator - thanks for your questions. We're discussing with the host and we'll post our decision as soon as it's made.
", "votes": null, "replies": [] }, { "id": 1272660, "author_name": "maggiemd", "author_url": "", "post_date": "04/13/2021 17:04:34", "content": "We created a post to capture external data sources. External data is allowed but it must be freely and publicly available and must be sited before the competition deadline.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1270629": "There may be ways to improve prediction quality by training our models using data beyond the given training set. First choice, naturally, may be to search the sites of the organizations mentioned in the [competition overview](https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/overview) like USDA, NSF, CHORUS, NIH etc. - thanks to the organizers for this hint. But there surely are many more sources out there. Besides, external data is indirectly used by those of us who import pre-trained models.\nIt may help to establish guidelines on use and disclosure in this respect to avoid surprises towards the end of the competition. Could @julialane or someone else in the organizers team clarify the external data use aspects.\n- is there any restriction on what external data may be used for model training?\n- what kind of disclosure is required on external data sources (none, source name reported in certain thread, full dataset loaded to Kaggle)? \n- shall we disclose use of the data provided by the organizations, mentioned by the organizers in the competition overview, or they are considered 'pre-approved'?\n- what are the time restrictions if any in declaring used external sources?", "1271670": "Hi @poedator - thanks for your questions. We're discussing with the host and we'll post our decision as soon as it's made.", "1272660": "We created a post to capture[ external data sources](https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/232414). External data is allowed but it must be freely and publicly available and must be sited before the competition deadline." }, "source": "meta" }