Probably a dumb question but I just want to be sure.
\nWill the datasets included in the real test set be the same as those present in the training set ? I guess the answer to this question will define if the problem at hand is a supervised or unsupervised one.
", "messages": [ { "id": 1253068, "postDate": "2021-03-26T10:40:54.397Z", "content": "Probably a dumb question but I just want to be sure.
\nWill the datasets included in the real test set be the same as those present in the training set ? I guess the answer to this question will define if the problem at hand is a supervised or unsupervised one.
", "rawMarkdown": "Probably a dumb question but I just want to be sure. \n\nWill the datasets included in the real test set be the same as those present in the training set ? I guess the answer to this question will define if the problem at hand is a supervised or unsupervised one. \n", "votes": 3 }, { "id": 1261335, "postDate": "2021-04-03T01:00:46.120Z", "content": "@philculliton could you please explain about test data size, i can see only 4 ids for prediction. But in data page, its mentioned test publication size is 8000. And when we get more data for test submission.
\nLeaderboard page, its mentioned 12% that is 960 records. Please clarify. We are not submitting our model, so how based on 4 publications - 960 predictions are happening.
", "rawMarkdown": "@philculliton could you please explain about test data size, i can see only 4 ids for prediction. But in data page, its mentioned test publication size is 8000. And when we get more data for test submission.\n\nLeaderboard page, its mentioned 12% that is 960 records. Please clarify. We are not submitting our model, so how based on 4 publications - 960 predictions are happening.", "votes": 1, "replies": [ { "id": 1263592, "postDate": "2021-04-05T14:23:13.180Z", "content": "Hi! Good question. The 4 publications provided are just example data. When you submit your notebook, your code is run on the hidden test sets (both public and private) which, respectively, are roughly 960 and 7200 samples.
", "rawMarkdown": "Hi! Good question. The 4 publications provided are just example data. When you submit your notebook, your code is run on the hidden test sets (both public and private) which, respectively, are roughly 960 and 7200 samples.", "votes": 2 } ] }, { "id": 1256034, "postDate": "2021-03-29T13:37:52.373Z", "content": "Hi! Good question. No - there are a huge number of datasets in the test set that are not present in the training set. The idea is definitely not simply to match the datasets from the training set, but rather to build a model that looks for semantic / NLP context around those datasets and extrapolates it to unknown datasets. The training set is meant as a guide to the task.
", "rawMarkdown": "Hi! Good question. No - there are a huge number of datasets in the test set that are not present in the training set. The idea is definitely not simply to match the datasets from the training set, but rather to build a model that looks for semantic / NLP context around those datasets and extrapolates it to unknown datasets. The training set is meant as a guide to the task.", "votes": 1 }, { "id": 1261359, "postDate": "2021-04-03T02:09:11.697Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1261335, "author_name": "rajpur", "author_url": "", "post_date": "2021-04-03T01:00:46.120000", "content": "@philculliton could you please explain about test data size, i can see only 4 ids for prediction. But in data page, its mentioned test publication size is 8000. And when we get more data for test submission.
\nLeaderboard page, its mentioned 12% that is 960 records. Please clarify. We are not submitting our model, so how based on 4 publications - 960 predictions are happening.
", "votes": 1, "replies": [ { "id": 1263592, "author_name": "Phil Culliton", "author_url": "", "post_date": "2021-04-05T14:23:13.180000", "content": "Hi! Good question. The 4 publications provided are just example data. When you submit your notebook, your code is run on the hidden test sets (both public and private) which, respectively, are roughly 960 and 7200 samples.
", "votes": 2, "replies": [] } ] }, { "id": 1256034, "author_name": "Phil Culliton", "author_url": "", "post_date": "2021-03-29T13:37:52.373000", "content": "Hi! Good question. No - there are a huge number of datasets in the test set that are not present in the training set. The idea is definitely not simply to match the datasets from the training set, but rather to build a model that looks for semantic / NLP context around those datasets and extrapolates it to unknown datasets. The training set is meant as a guide to the task.
", "votes": 1, "replies": [] }, { "id": 1261359, "author_name": "", "author_url": "", "post_date": "2021-04-03T02:09:11.697000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1253068": "Probably a dumb question but I just want to be sure. \n\nWill the datasets included in the real test set be the same as those present in the training set ? I guess the answer to this question will define if the problem at hand is a supervised or unsupervised one. \n", "1261335": "@philculliton could you please explain about test data size, i can see only 4 ids for prediction. But in data page, its mentioned test publication size is 8000. And when we get more data for test submission.\n\nLeaderboard page, its mentioned 12% that is 960 records. Please clarify. We are not submitting our model, so how based on 4 publications - 960 predictions are happening.", "1256034": "Hi! Good question. No - there are a huge number of datasets in the test set that are not present in the training set. The idea is definitely not simply to match the datasets from the training set, but rather to build a model that looks for semantic / NLP context around those datasets and extrapolates it to unknown datasets. The training set is meant as a guide to the task.", "1261359": "" } }