Just checking if I got it right.
\n\"It's worth noting that some external data providers might not be available for future (test) evaluations, as anticipated.\" Does this imply that certain columns, tables, or both, from external data, could be unavailable during future test evaluations? @tomasjeline2 and @jetakow
", "messages": [ { "id": 2659511, "postDate": "2024-02-19T22:41:48.313Z", "content": "Just checking if I got it right.
\n\"It's worth noting that some external data providers might not be available for future (test) evaluations, as anticipated.\" Does this imply that certain columns, tables, or both, from external data, could be unavailable during future test evaluations? @tomasjeline2 and @jetakow
", "rawMarkdown": "Just checking if I got it right. \n\n\"It's worth noting that some external data providers might not be available for future (test) evaluations, as anticipated.\" Does this imply that certain columns, tables, or both, from external data, could be unavailable during future test evaluations? @tomasjeline2 and @jetakow ", "votes": 2 }, { "id": 2659921, "postDate": "2024-02-20T08:05:44.560Z", "content": "Hi Bolin,
\nYes, it's possible.
\nLet me briefly why: it's not that we don't want to share the data, or that we want to create some artificial obstacle to achieve stability. But simply the external source became unavailable to us - provider lost access to data, quality of data became unsatisfactory, can't provide data due to privacy issues, disagreement on pricing, or alternative data became available.
That makes sense. In the further evaluation stage, if columns or tables are unavailable, would you still provide them with the same names as used in the external training data but filled with NaN?
", "rawMarkdown": "That makes sense. In the further evaluation stage, if columns or tables are unavailable, would you still provide them with the same names as used in the external training data but filled with NaN?", "replies": [ { "id": 2660147, "postDate": "2024-02-20T11:32:03.523Z", "content": "yes, structure of tables remains same, but with null/NaN values
", "rawMarkdown": "yes, structure of tables remains same, but with null/NaN values" }, { "id": 2660160, "postDate": "2024-02-20T11:50:50.603Z", "content": "Great. That is very helpful.
", "rawMarkdown": "Great. That is very helpful." } ] }, { "id": 2753555, "postDate": "2024-04-15T15:07:16.100Z", "content": "thanks Tomas and Bolin.
\na follow up question: in the case of future missing external data, will the number of files under one data source change?
\nfor example, in private test data set, if all test_credit_bureau_a_2_* data are missing, will we still have 11 files or only 1 file named 'test_credit_bureau_a_2_0' with NaN under all columns?
", "rawMarkdown": "thanks Tomas and Bolin. \n\na follow up question: in the case of future missing external data, will the number of files under one data source change?\n\nfor example, in private test data set, if all test_credit_bureau_a_2_* data are missing, will we still have 11 files or only 1 file named 'test_credit_bureau_a_2_0' with NaN under all columns?" } ] } ], "comments": [ { "id": 2659921, "author_name": "Tomas Jelinek", "author_url": "", "post_date": "2024-02-20T08:05:44.560000", "content": "Hi Bolin,
\nYes, it's possible.
\nLet me briefly why: it's not that we don't want to share the data, or that we want to create some artificial obstacle to achieve stability. But simply the external source became unavailable to us - provider lost access to data, quality of data became unsatisfactory, can't provide data due to privacy issues, disagreement on pricing, or alternative data became available.
That makes sense. In the further evaluation stage, if columns or tables are unavailable, would you still provide them with the same names as used in the external training data but filled with NaN?
", "votes": 0, "replies": [ { "id": 2660147, "author_name": "Tomas Jelinek", "author_url": "", "post_date": "2024-02-20T11:32:03.523000", "content": "yes, structure of tables remains same, but with null/NaN values
", "votes": 0, "replies": [] }, { "id": 2660160, "author_name": "Bolin Pan", "author_url": "", "post_date": "2024-02-20T11:50:50.603000", "content": "Great. That is very helpful.
", "votes": 0, "replies": [] } ] }, { "id": 2753555, "author_name": "Shiyu Du", "author_url": "", "post_date": "2024-04-15T15:07:16.100000", "content": "thanks Tomas and Bolin.
\na follow up question: in the case of future missing external data, will the number of files under one data source change?
\nfor example, in private test data set, if all test_credit_bureau_a_2_* data are missing, will we still have 11 files or only 1 file named 'test_credit_bureau_a_2_0' with NaN under all columns?
", "votes": 0, "replies": [] } ] } ], "raw_markdown_by_id": { "2659511": "Just checking if I got it right. \n\n\"It's worth noting that some external data providers might not be available for future (test) evaluations, as anticipated.\" Does this imply that certain columns, tables, or both, from external data, could be unavailable during future test evaluations? @tomasjeline2 and @jetakow ", "2659921": "Hi Bolin,\nYes, it's possible.\nLet me briefly why: it's not that we don't want to share the data, or that we want to create some artificial obstacle to achieve stability. But simply the external source became unavailable to us - provider lost access to data, quality of data became unsatisfactory, can't provide data due to privacy issues, disagreement on pricing, or alternative data became available." } }