{
  "id": 476589,
  "title": "NaN values at eeg",
  "url": "/competitions/hms-harmful-brain-activity-classification/discussion/476589",
  "author_name": "",
  "post_date": "2024-02-12T23:20:00.307583600Z",
  "votes": 6,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi. I've been visually inspecting the nan values on eeg and I've found that they're in my opinion either of two causes:</p>\n<ol>\n<li>Puntual miss readings of 1/200 seconds that can be simply ignored or linearly interpolated.</li>\n<li>Just initial or ending paddings.</li>\n</ol>\n<p>I'm right? Have someone found different nan values to take care off?</p>\n<p>EDIT: Just finished a complete train/EEG visually inspection. I though it would take longer. The higher non initial non ending nan gap was of seven non continuous 1/200s. So yes. I think I was right.</p>",
  "messages": [
    {
      "id": "2649538",
      "postDate": "02/12/2024 23:20:00",
      "content": "<p>Hi. I've been visually inspecting the nan values on eeg and I've found that they're in my opinion either of two causes:</p>\n<ol>\n<li>Puntual miss readings of 1/200 seconds that can be simply ignored or linearly interpolated.</li>\n<li>Just initial or ending paddings.</li>\n</ol>\n<p>I'm right? Have someone found different nan values to take care off?</p>\n<p>EDIT: Just finished a complete train/EEG visually inspection. I though it would take longer. The higher non initial non ending nan gap was of seven non continuous 1/200s. So yes. I think I was right.</p>",
      "rawMarkdown": "Hi. I've been visually inspecting the nan values on eeg and I've found that they're in my opinion either of two causes:\n1. Puntual miss readings of 1/200 seconds that can be simply ignored or linearly interpolated.\n2. Just initial or ending paddings.\n\nI'm right? Have someone found different nan values to take care off?\n\nEDIT: Just finished a complete train/EEG visually inspection. I though it would take longer. The higher non initial non ending nan gap was of seven non continuous 1/200s. So yes. I think I was right.",
      "votes": null
    },
    {
      "id": "2653886",
      "postDate": "02/15/2024 17:30:22",
      "content": "<p>I found around ~850 eeg signals with nans, most of them have 40-80 nans, but their is ~85 eeg signals that are <em>mostly nans</em>. You can load all the eeg ids and their nan count from <a href=\"https://www.kaggle.com/datasets/nartaa/eeg-spectrograms?select=eeg_nans.npy\" target=\"_blank\">here</a>.</p>",
      "rawMarkdown": "I found around ~850 eeg signals with nans, most of them have 40-80 nans, but their is ~85 eeg signals that are *mostly nans*. You can load all the eeg ids and their nan count from [here](https://www.kaggle.com/datasets/nartaa/eeg-spectrograms?select=eeg_nans.npy).",
      "votes": null
    },
    {
      "id": "2653936",
      "postDate": "02/15/2024 18:22:57",
      "content": "<p>\"most of them have 40-80 nans\"<br>\nThere is 20 electrodes. Those nans corresponds to 2-4 non continuous 1/200s gaps. First case.<br>\n\"but their is ~85 eeg signals that are mostly nans\"<br>\nThose nans corresponds to initial or ending continuos gaps on time dimension. Like a padding. 2nd case. You can check it.</p>",
      "rawMarkdown": "\"most of them have 40-80 nans\"\nThere is 20 electrodes. Those nans corresponds to 2-4 non continuous 1/200s gaps. First case.\n\"but their is ~85 eeg signals that are mostly nans\"\nThose nans corresponds to initial or ending continuos gaps on time dimension. Like a padding. 2nd case. You can check it.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2653886,
      "author_name": "nartaa",
      "author_url": "",
      "post_date": "02/15/2024 17:30:22",
      "content": "<p>I found around ~850 eeg signals with nans, most of them have 40-80 nans, but their is ~85 eeg signals that are <em>mostly nans</em>. You can load all the eeg ids and their nan count from <a href=\"https://www.kaggle.com/datasets/nartaa/eeg-spectrograms?select=eeg_nans.npy\" target=\"_blank\">here</a>.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2653936,
          "author_name": "sacuscreed",
          "author_url": "",
          "post_date": "02/15/2024 18:22:57",
          "content": "<p>\"most of them have 40-80 nans\"<br>\nThere is 20 electrodes. Those nans corresponds to 2-4 non continuous 1/200s gaps. First case.<br>\n\"but their is ~85 eeg signals that are mostly nans\"<br>\nThose nans corresponds to initial or ending continuos gaps on time dimension. Like a padding. 2nd case. You can check it.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2649538": "Hi. I've been visually inspecting the nan values on eeg and I've found that they're in my opinion either of two causes:\n1. Puntual miss readings of 1/200 seconds that can be simply ignored or linearly interpolated.\n2. Just initial or ending paddings.\n\nI'm right? Have someone found different nan values to take care off?\n\nEDIT: Just finished a complete train/EEG visually inspection. I though it would take longer. The higher non initial non ending nan gap was of seven non continuous 1/200s. So yes. I think I was right.",
    "2653886": "I found around ~850 eeg signals with nans, most of them have 40-80 nans, but their is ~85 eeg signals that are *mostly nans*. You can load all the eeg ids and their nan count from [here](https://www.kaggle.com/datasets/nartaa/eeg-spectrograms?select=eeg_nans.npy).",
    "2653936": "\"most of them have 40-80 nans\"\nThere is 20 electrodes. Those nans corresponds to 2-4 non continuous 1/200s gaps. First case.\n\"but their is ~85 eeg signals that are mostly nans\"\nThose nans corresponds to initial or ending continuos gaps on time dimension. Like a padding. 2nd case. You can check it."
  },
  "source": "meta"
}