Are not the study labels for classification mutually exclusive？
", "messages": [ { "id": "1323664", "postDate": "05/26/2021 11:42:25", "content": "Are not the study labels for classification mutually exclusive？
", "rawMarkdown": "Are not the study labels for classification mutually exclusive？", "votes": null }, { "id": "1324071", "postDate": "05/26/2021 15:39:11", "content": "Yes, study labels are mutually exclusive even though its mentioned \"Studies in the test set may contain more than one label.\" in Evaluation page. There is also mention to mutually exclusiveness here. In the training data each study has only 1 label. Although, this information alone doesn't guarantee this will be the case for the hidden test set all the public information provided from the host supports mutually exclusive study labels.
Also, most of the studies have 1 image and since a radiograph/image can have 1 of the 4 labels study will also have only 1 label. There are studies with multiple images but those images are mostly augmented versions of another image from the same study. By taking all of the information into account it is easier to assume that a study is labelled with either one of the \"negative\", \"typical\", \"indeterminate\", \"atypical\" labels.
In my thread I got https://www.kaggle.com/c/siim-covid19-detection/discussion/241238 response from Phil from Kaggle \"all studies should have one study-level label.\"
", "rawMarkdown": "In my thread I got https://www.kaggle.com/c/siim-covid19-detection/discussion/241238 response from Phil from Kaggle \"all studies should have one study-level label.\"", "votes": null } ], "comments": [ { "id": 1324071, "author_name": "keremt", "author_url": "", "post_date": "05/26/2021 15:39:11", "content": "Yes, study labels are mutually exclusive even though its mentioned \"Studies in the test set may contain more than one label.\" in Evaluation page. There is also mention to mutually exclusiveness here. In the training data each study has only 1 label. Although, this information alone doesn't guarantee this will be the case for the hidden test set all the public information provided from the host supports mutually exclusive study labels.
Also, most of the studies have 1 image and since a radiograph/image can have 1 of the 4 labels study will also have only 1 label. There are studies with multiple images but those images are mostly augmented versions of another image from the same study. By taking all of the information into account it is easier to assume that a study is labelled with either one of the \"negative\", \"typical\", \"indeterminate\", \"atypical\" labels.
In my thread I got https://www.kaggle.com/c/siim-covid19-detection/discussion/241238 response from Phil from Kaggle \"all studies should have one study-level label.\"
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1323664": "Are not the study labels for classification mutually exclusive？", "1324071": "Yes, study labels are mutually exclusive even though its mentioned `\"Studies in the test set may contain more than one label.\"` in [Evaluation](https://www.kaggle.com/c/siim-covid19-detection/overview/evaluation) page. There is also mention to mutually exclusiveness [here](https://www.kaggle.com/c/siim-covid19-detection/discussion/240250). In the training data each study has only 1 label. Although, this information alone doesn't guarantee this will be the case for the hidden test set all the public information provided from the host supports mutually exclusive study labels.\n\nAlso, most of the studies have 1 image and since a radiograph/image can have 1 of the 4 labels study will also have only 1 label. There are studies with multiple images but those images are mostly augmented versions of another image from the same study. By taking all of the information into account it is easier to assume that a study is labelled with either one of the `\"negative\", \"typical\", \"indeterminate\", \"atypical\"` labels.", "1324236": "In my thread I got https://www.kaggle.com/c/siim-covid19-detection/discussion/241238 response from Phil from Kaggle \"all studies should have one study-level label.\"" }, "source": "meta" }