{
  "id": 582597,
  "title": "Full resolution data vs compressed data performance difference?",
  "url": "/competitions/waveform-inversion/discussion/582597",
  "author_name": "",
  "post_date": "2025-06-01T11:05:35.767944200Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p><a href=\"https://www.kaggle.com/datasets/brendanartley/openfwi-preprocessed-72x72/data\" target=\"_blank\">https://www.kaggle.com/datasets/brendanartley/openfwi-preprocessed-72x72/data</a></p>\n<p>Someone released a compressed dataset which is only around 4% of the original dataset.</p>\n<p>How well do the models perform trained on the compressed data in comparison to the original data?</p>",
  "messages": [
    {
      "id": "3214970",
      "postDate": "06/01/2025 11:05:35",
      "content": "<p><a href=\"https://www.kaggle.com/datasets/brendanartley/openfwi-preprocessed-72x72/data\" target=\"_blank\">https://www.kaggle.com/datasets/brendanartley/openfwi-preprocessed-72x72/data</a></p>\n<p>Someone released a compressed dataset which is only around 4% of the original dataset.</p>\n<p>How well do the models perform trained on the compressed data in comparison to the original data?</p>",
      "rawMarkdown": "https://www.kaggle.com/datasets/brendanartley/openfwi-preprocessed-72x72/data\n\nSomeone released a compressed dataset which is only around 4% of the original dataset.\n\nHow well do the models perform trained on the compressed data in comparison to the original data?",
      "votes": null
    },
    {
      "id": "3215020",
      "postDate": "06/01/2025 13:15:30",
      "content": "<p>around 30 mae</p>",
      "rawMarkdown": "around 30 mae",
      "votes": null
    },
    {
      "id": "3215294",
      "postDate": "06/02/2025 01:17:19",
      "content": "<p>That's better than I expected, thanks. But how does it compare to the original full resolution data?</p>\n<p>For example, what mae would you get if you train the same model on the original data?</p>",
      "rawMarkdown": "That's better than I expected, thanks. But how does it compare to the original full resolution data?\n\nFor example, what mae would you get if you train the same model on the original data?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3215020,
      "author_name": "aichangeworld",
      "author_url": "",
      "post_date": "06/01/2025 13:15:30",
      "content": "<p>around 30 mae</p>",
      "votes": null,
      "replies": [
        {
          "id": 3215294,
          "author_name": "max6296",
          "author_url": "",
          "post_date": "06/02/2025 01:17:19",
          "content": "<p>That's better than I expected, thanks. But how does it compare to the original full resolution data?</p>\n<p>For example, what mae would you get if you train the same model on the original data?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3214970": "https://www.kaggle.com/datasets/brendanartley/openfwi-preprocessed-72x72/data\n\nSomeone released a compressed dataset which is only around 4% of the original dataset.\n\nHow well do the models perform trained on the compressed data in comparison to the original data?",
    "3215020": "around 30 mae",
    "3215294": "That's better than I expected, thanks. But how does it compare to the original full resolution data?\n\nFor example, what mae would you get if you train the same model on the original data?"
  },
  "source": "meta"
}