{
  "id": 471446,
  "title": "Author Paper: Development of Expert-Level Classification of Seizures and Rhythmic and Periodic Patterns During EEG Interpretation",
  "url": "/competitions/hms-harmful-brain-activity-classification/discussion/471446",
  "author_name": "",
  "post_date": "2024-01-28T11:14:04.858779500Z",
  "votes": 18,
  "comment_count": 1,
  "views": 0,
  "content": "<h1>From the <a href=\"https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/471439\" target=\"_blank\">shared discussion</a></h1>\n<h1>How experts see the data</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F63efb384629a0630577593d1dcffcf38%2FScreenshot%202024-01-28%20at%204.34.26PM.png?generation=1706439883645021&amp;alt=media\"></p>\n<blockquote>\n  <h3>==&gt; 10secs sub-sequence of eeg and 4 spectrograms and clustering ( what use of cluster umap ? )</h3>\n</blockquote>\n<hr>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F6bd53f841cc12aa2735bdee387c8b311%2FScreenshot%202024-01-28%20at%205.53.54PM.png?generation=1706444650452211&amp;alt=media\"></p>\n<blockquote>\n  <h3>==&gt; Only Specific bands are useful to classify respective targets ( can be used for feature engineering - but why 0 to 20 ? not to 0 to 40 KHz )</h3>\n</blockquote>\n<hr>\n<h1>Data labels (maybe similar to competition dataset)</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Fbb771a8a4ecdcfb97eae9f42984b2e5c%2FScreenshot%202024-01-28%20at%204.36.48PM.png?generation=1706440028960948&amp;alt=media\"></p>\n<hr>\n<h1>CV Split ( need to be tested )</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F5b1d79f8331eb3205edbc117b891d503%2FScreenshot%202024-01-28%20at%204.38.24PM.png?generation=1706440116935386&amp;alt=media\"></p>\n<hr>\n<h1>Model Architecture</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F835d55f50177a9cf193f2f89473e950b%2FScreenshot%202024-01-28%20at%204.38.56PM.png?generation=1706440152800415&amp;alt=media\"></p>\n<hr>\n<h1>Data Augmentation</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Faf57daa8b2c74ffd715bdb1800add8d9%2FScreenshot%202024-01-28%20at%204.41.57PM.png?generation=1706440331073506&amp;alt=media\"></p>\n<hr>\n<h1>Post threshold ??</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Fc947de0a8e582d9385ac06815df1e838%2FScreenshot%202024-01-28%20at%204.39.52PM.png?generation=1706440223958195&amp;alt=media\"></p>",
  "messages": [
    {
      "id": "2623776",
      "postDate": "01/28/2024 11:14:04",
      "content": "<h1>From the <a href=\"https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/471439\" target=\"_blank\">shared discussion</a></h1>\n<h1>How experts see the data</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F63efb384629a0630577593d1dcffcf38%2FScreenshot%202024-01-28%20at%204.34.26PM.png?generation=1706439883645021&amp;alt=media\"></p>\n<blockquote>\n  <h3>==&gt; 10secs sub-sequence of eeg and 4 spectrograms and clustering ( what use of cluster umap ? )</h3>\n</blockquote>\n<hr>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F6bd53f841cc12aa2735bdee387c8b311%2FScreenshot%202024-01-28%20at%205.53.54PM.png?generation=1706444650452211&amp;alt=media\"></p>\n<blockquote>\n  <h3>==&gt; Only Specific bands are useful to classify respective targets ( can be used for feature engineering - but why 0 to 20 ? not to 0 to 40 KHz )</h3>\n</blockquote>\n<hr>\n<h1>Data labels (maybe similar to competition dataset)</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Fbb771a8a4ecdcfb97eae9f42984b2e5c%2FScreenshot%202024-01-28%20at%204.36.48PM.png?generation=1706440028960948&amp;alt=media\"></p>\n<hr>\n<h1>CV Split ( need to be tested )</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F5b1d79f8331eb3205edbc117b891d503%2FScreenshot%202024-01-28%20at%204.38.24PM.png?generation=1706440116935386&amp;alt=media\"></p>\n<hr>\n<h1>Model Architecture</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F835d55f50177a9cf193f2f89473e950b%2FScreenshot%202024-01-28%20at%204.38.56PM.png?generation=1706440152800415&amp;alt=media\"></p>\n<hr>\n<h1>Data Augmentation</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Faf57daa8b2c74ffd715bdb1800add8d9%2FScreenshot%202024-01-28%20at%204.41.57PM.png?generation=1706440331073506&amp;alt=media\"></p>\n<hr>\n<h1>Post threshold ??</h1>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Fc947de0a8e582d9385ac06815df1e838%2FScreenshot%202024-01-28%20at%204.39.52PM.png?generation=1706440223958195&amp;alt=media\"></p>",
      "rawMarkdown": "# From the [shared discussion](https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/471439)\n\n# How experts see the data\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F63efb384629a0630577593d1dcffcf38%2FScreenshot%202024-01-28%20at%204.34.26PM.png?generation=1706439883645021&alt=media)\n> ### ==> 10secs sub-sequence of eeg and 4 spectrograms and clustering ( what use of cluster umap ? )\n\n---\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F6bd53f841cc12aa2735bdee387c8b311%2FScreenshot%202024-01-28%20at%205.53.54PM.png?generation=1706444650452211&alt=media)\n> ### ==> Only Specific bands are useful to classify respective targets ( can be used for feature engineering - but why 0 to 20 ? not to 0 to 40 KHz )\n\n---\n# Data labels (maybe similar to competition dataset)\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Fbb771a8a4ecdcfb97eae9f42984b2e5c%2FScreenshot%202024-01-28%20at%204.36.48PM.png?generation=1706440028960948&alt=media)\n\n---\n# CV Split ( need to be tested )\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F5b1d79f8331eb3205edbc117b891d503%2FScreenshot%202024-01-28%20at%204.38.24PM.png?generation=1706440116935386&alt=media)\n\n---\n# Model Architecture\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F835d55f50177a9cf193f2f89473e950b%2FScreenshot%202024-01-28%20at%204.38.56PM.png?generation=1706440152800415&alt=media)\n\n---\n# Data Augmentation\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Faf57daa8b2c74ffd715bdb1800add8d9%2FScreenshot%202024-01-28%20at%204.41.57PM.png?generation=1706440331073506&alt=media)\n\n---\n# Post threshold ?? \n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Fc947de0a8e582d9385ac06815df1e838%2FScreenshot%202024-01-28%20at%204.39.52PM.png?generation=1706440223958195&alt=media)",
      "votes": null
    },
    {
      "id": "2633214",
      "postDate": "02/02/2024 21:47:19",
      "content": "<p>Very helpful!  Interesting analysis of why your data augmentation results in useful data.  Thanks, Matthew</p>",
      "rawMarkdown": "Very helpful!  Interesting analysis of why your data augmentation results in useful data.  Thanks, Matthew",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2633214,
      "author_name": "matucker",
      "author_url": "",
      "post_date": "02/02/2024 21:47:19",
      "content": "<p>Very helpful!  Interesting analysis of why your data augmentation results in useful data.  Thanks, Matthew</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2623776": "# From the [shared discussion](https://www.kaggle.com/competitions/hms-harmful-brain-activity-classification/discussion/471439)\n\n# How experts see the data\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F63efb384629a0630577593d1dcffcf38%2FScreenshot%202024-01-28%20at%204.34.26PM.png?generation=1706439883645021&alt=media)\n> ### ==> 10secs sub-sequence of eeg and 4 spectrograms and clustering ( what use of cluster umap ? )\n\n---\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F6bd53f841cc12aa2735bdee387c8b311%2FScreenshot%202024-01-28%20at%205.53.54PM.png?generation=1706444650452211&alt=media)\n> ### ==> Only Specific bands are useful to classify respective targets ( can be used for feature engineering - but why 0 to 20 ? not to 0 to 40 KHz )\n\n---\n# Data labels (maybe similar to competition dataset)\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Fbb771a8a4ecdcfb97eae9f42984b2e5c%2FScreenshot%202024-01-28%20at%204.36.48PM.png?generation=1706440028960948&alt=media)\n\n---\n# CV Split ( need to be tested )\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F5b1d79f8331eb3205edbc117b891d503%2FScreenshot%202024-01-28%20at%204.38.24PM.png?generation=1706440116935386&alt=media)\n\n---\n# Model Architecture\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2F835d55f50177a9cf193f2f89473e950b%2FScreenshot%202024-01-28%20at%204.38.56PM.png?generation=1706440152800415&alt=media)\n\n---\n# Data Augmentation\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Faf57daa8b2c74ffd715bdb1800add8d9%2FScreenshot%202024-01-28%20at%204.41.57PM.png?generation=1706440331073506&alt=media)\n\n---\n# Post threshold ?? \n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F761268%2Fc947de0a8e582d9385ac06815df1e838%2FScreenshot%202024-01-28%20at%204.39.52PM.png?generation=1706440223958195&alt=media)",
    "2633214": "Very helpful!  Interesting analysis of why your data augmentation results in useful data.  Thanks, Matthew"
  },
  "source": "meta"
}