For example, for the same patient_id, maybe have several rows in train.csv, which have different XXX_sub_id but the same expert_concensus, i don't know what's the difference between them? And because of that, when we use all of that to build a dataset for training model, that might cause overfitting during the training process because there are a lot of overlapping in the dataset! Any sharing would be appreciate!
", "messages": [ { "id": "2607578", "postDate": "01/18/2024 10:15:47", "content": "For example, for the same patient_id, maybe have several rows in train.csv, which have different XXX_sub_id but the same expert_concensus, i don't know what's the difference between them? And because of that, when we use all of that to build a dataset for training model, that might cause overfitting during the training process because there are a lot of overlapping in the dataset! Any sharing would be appreciate!
", "rawMarkdown": "For example, for the same patient_id, maybe have several rows in train.csv, which have different XXX_sub_id but the same expert_concensus, i don't know what's the difference between them? And because of that, when we use all of that to build a dataset for training model, that might cause overfitting during the training process because there are a lot of overlapping in the dataset! Any sharing would be appreciate!", "votes": null }, { "id": "2607856", "postDate": "01/18/2024 13:34:42", "content": "@suanbaicai go through this discussion & discussion
", "rawMarkdown": "suanbaicai go through this [discussion](https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/468208) & [discussion](https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/468010)", "votes": null }, { "id": "2609234", "postDate": "01/19/2024 10:31:45", "content": "First of all, Thanks for the reply!
\nYes, I've seen these two discussion, but two point still remain confused:
\n1.how does the label_id comes out, for example, for eeg_id=1628180742&eeg_sub_id=0, we can get label_id=127492639, why 127492639? not 3.1415926 or sth else. Is it the time location of the 10-sec signal or sth else?
\n2.why just use 10-sec signal in the middle to predict, i learned the fact throught the discussion, but don't know why
how does the label_id comes out, for example, for eeg_id=1628180742&eeg_sub_id=0, we can get label_id=127492639, why 127492639? not 3.1415926 or sth else. Is it the time location of the 10-sec signal or sth else?
\n\n\nlabel_id not sure, but # of label_id is same as # of rows in train.csv ( mostly my guess is label_id is unique for the 10secs seq pattern )
\n
why just use 10-sec signal in the middle to predict, i learned the fact throught the discussion, but don't know why
\n\n", "rawMarkdown": "how does the label_id comes out, for example, for eeg_id=1628180742&eeg_sub_id=0, we can get label_id=127492639, why 127492639? not 3.1415926 or sth else. Is it the time location of the 10-sec signal or sth else?\n\n> label_id not sure, but # of label_id is same as # of rows in train.csv ( mostly my guess is label_id is unique for the 10secs seq pattern )\n\nwhy just use 10-sec signal in the middle to predict, i learned the fact throught the discussion, but don't know why\n\n> @suanbaicai Experts focus on 10-sec while voting the targets even they see 50secs of eeg", "votes": null } ], "comments": [ { "id": 2607856, "author_name": "seshurajup", "author_url": "", "post_date": "01/18/2024 13:34:42", "content": "@suanbaicai Experts focus on 10-sec while voting the targets even they see 50secs of eeg
\n
@suanbaicai go through this discussion & discussion
", "votes": null, "replies": [ { "id": 2609234, "author_name": "suanbaicai", "author_url": "", "post_date": "01/19/2024 10:31:45", "content": "First of all, Thanks for the reply!
\nYes, I've seen these two discussion, but two point still remain confused:
\n1.how does the label_id comes out, for example, for eeg_id=1628180742&eeg_sub_id=0, we can get label_id=127492639, why 127492639? not 3.1415926 or sth else. Is it the time location of the 10-sec signal or sth else?
\n2.why just use 10-sec signal in the middle to predict, i learned the fact throught the discussion, but don't know why
how does the label_id comes out, for example, for eeg_id=1628180742&eeg_sub_id=0, we can get label_id=127492639, why 127492639? not 3.1415926 or sth else. Is it the time location of the 10-sec signal or sth else?
\n\n\nlabel_id not sure, but # of label_id is same as # of rows in train.csv ( mostly my guess is label_id is unique for the 10secs seq pattern )
\n
why just use 10-sec signal in the middle to predict, i learned the fact throught the discussion, but don't know why
\n\n", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2607578": "For example, for the same patient_id, maybe have several rows in train.csv, which have different XXX_sub_id but the same expert_concensus, i don't know what's the difference between them? And because of that, when we use all of that to build a dataset for training model, that might cause overfitting during the training process because there are a lot of overlapping in the dataset! Any sharing would be appreciate!", "2607856": "suanbaicai go through this [discussion](https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/468208) & [discussion](https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/468010)", "2609234": "First of all, Thanks for the reply!\nYes, I've seen these two discussion, but two point still remain confused:\n1.how does the label_id comes out, for example, for eeg_id=1628180742&eeg_sub_id=0, we can get label_id=127492639, why 127492639? not 3.1415926 or sth else. Is it the time location of the 10-sec signal or sth else?\n2.why just use 10-sec signal in the middle to predict, i learned the fact throught the discussion, but don't know why", "2610216": "how does the label_id comes out, for example, for eeg_id=1628180742&eeg_sub_id=0, we can get label_id=127492639, why 127492639? not 3.1415926 or sth else. Is it the time location of the 10-sec signal or sth else?\n\n> label_id not sure, but # of label_id is same as # of rows in train.csv ( mostly my guess is label_id is unique for the 10secs seq pattern )\n\nwhy just use 10-sec signal in the middle to predict, i learned the fact throught the discussion, but don't know why\n\n> @suanbaicai Experts focus on 10-sec while voting the targets even they see 50secs of eeg" }, "source": "meta" }@suanbaicai Experts focus on 10-sec while voting the targets even they see 50secs of eeg
\n