{
  "id": 583563,
  "title": "Physics informed Loss?",
  "url": "/competitions/waveform-inversion/discussion/583563",
  "author_name": "",
  "post_date": "2025-06-07T20:48:46.407198600Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Anyone has experimented with physics informed Loss or L1 loss is still considered good for this data?</p>",
  "messages": [
    {
      "id": "3219511",
      "postDate": "06/07/2025 20:48:46",
      "content": "<p>Anyone has experimented with physics informed Loss or L1 loss is still considered good for this data?</p>",
      "rawMarkdown": "Anyone has experimented with physics informed Loss or L1 loss is still considered good for this data?",
      "votes": null
    },
    {
      "id": "3219532",
      "postDate": "06/07/2025 22:35:29",
      "content": "<p>I was using Physics informed loss, and to be honest didnt like it too much. I started to try a loss function that combines MAE (primary) with gradient preservation (0.1 weight) and smoothness regularization (0.05 weight).</p>",
      "rawMarkdown": "I was using Physics informed loss, and to be honest didnt like it too much. I started to try a loss function that combines MAE (primary) with gradient preservation (0.1 weight) and smoothness regularization (0.05 weight).",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3219532,
      "author_name": "alexandersassi",
      "author_url": "",
      "post_date": "06/07/2025 22:35:29",
      "content": "<p>I was using Physics informed loss, and to be honest didnt like it too much. I started to try a loss function that combines MAE (primary) with gradient preservation (0.1 weight) and smoothness regularization (0.05 weight).</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3219511": "Anyone has experimented with physics informed Loss or L1 loss is still considered good for this data?",
    "3219532": "I was using Physics informed loss, and to be honest didnt like it too much. I started to try a loss function that combines MAE (primary) with gradient preservation (0.1 weight) and smoothness regularization (0.05 weight)."
  },
  "source": "meta"
}