{
  "id": 542222,
  "title": "Wondering how to select the best features",
  "url": "/competitions/airs-ai-in-respiratory-sounds/discussion/542222",
  "author_name": "Himanshu Kaushik",
  "post_date": "2024-10-23T14:43:53.840000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>There a number of features available for for each data sample to be extracted. for e.g.</p>\n<ol>\n<li>Using a pretrained existing model to generate embeddings of cough sound as well as vowel sound</li>\n<li>Using biomarker embeddings given using google HeAR model</li>\n<li>Create new features from MFCC and other properties of sound files given</li>\n<li>The information given about the patients in the table.</li>\n</ol>\n<p>What should be the quickest and easiest approach to be tried first. It would be interesting to discuss how to decide this. what are your thoughts?</p>",
  "messages": [
    {
      "id": 3026213,
      "postDate": "2024-10-23T14:43:53.840Z",
      "content": "<p>There a number of features available for for each data sample to be extracted. for e.g.</p>\n<ol>\n<li>Using a pretrained existing model to generate embeddings of cough sound as well as vowel sound</li>\n<li>Using biomarker embeddings given using google HeAR model</li>\n<li>Create new features from MFCC and other properties of sound files given</li>\n<li>The information given about the patients in the table.</li>\n</ol>\n<p>What should be the quickest and easiest approach to be tried first. It would be interesting to discuss how to decide this. what are your thoughts?</p>",
      "rawMarkdown": "There a number of features available for for each data sample to be extracted. for e.g.\n1. Using a pretrained existing model to generate embeddings of cough sound as well as vowel sound\n2. Using biomarker embeddings given using google HeAR model\n3. Create new features from MFCC and other properties of sound files given\n4. The information given about the patients in the table.\n\nWhat should be the quickest and easiest approach to be tried first. It would be interesting to discuss how to decide this. what are your thoughts?"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3026213": "There a number of features available for for each data sample to be extracted. for e.g.\n1. Using a pretrained existing model to generate embeddings of cough sound as well as vowel sound\n2. Using biomarker embeddings given using google HeAR model\n3. Create new features from MFCC and other properties of sound files given\n4. The information given about the patients in the table.\n\nWhat should be the quickest and easiest approach to be tried first. It would be interesting to discuss how to decide this. what are your thoughts?"
  }
}