{
  "id": 573962,
  "title": "Data Augmentation with Noise during Training",
  "url": "/competitions/byu-locating-bacterial-flagellar-motors-2025/discussion/573962",
  "author_name": "Ali Kabirzadeh",
  "post_date": "2025-04-18T23:41:17.522000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Using weak noise led to a decrease in model performance, because it was too similar to the original training data and acted more like adding duplicate data rather than effective augmentation.</p>\n<p><a href=\"url\" target=\"_blank\">https://www.kaggle.com/code/ali78kabirzadeh/bactria-s2</a></p>",
  "messages": [
    {
      "id": 3182221,
      "postDate": "2025-04-18T23:41:17.523Z",
      "content": "<p>Using weak noise led to a decrease in model performance, because it was too similar to the original training data and acted more like adding duplicate data rather than effective augmentation.</p>\n<p><a href=\"url\" target=\"_blank\">https://www.kaggle.com/code/ali78kabirzadeh/bactria-s2</a></p>",
      "rawMarkdown": "Using weak noise led to a decrease in model performance, because it was too similar to the original training data and acted more like adding duplicate data rather than effective augmentation.\n\n[https://www.kaggle.com/code/ali78kabirzadeh/bactria-s2](url)",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3182221": "Using weak noise led to a decrease in model performance, because it was too similar to the original training data and acted more like adding duplicate data rather than effective augmentation.\n\n[https://www.kaggle.com/code/ali78kabirzadeh/bactria-s2](url)"
  }
}