{
  "id": 515652,
  "title": "Multiple models for each level or variable",
  "url": "/competitions/leap-atmospheric-physics-ai-climsim/discussion/515652",
  "author_name": "Vasilis",
  "post_date": "2024-06-29T08:13:40.448000",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hello guys, i have my main model and i decided to do an experiment and fork it to predict only for level 41 (one that i have the worst result). I let it train whole night and i was pretty sure i would see a small improvement in the loss of variables of level 41 but i did not. That was surprising for me, i also tried to predict all levels of one specific variable (this time i let it run only for a few iterations) and it did not seem promising. Anybody that tried this and got better result? Does it make sense that trying to predict much less targets did not produced better results?</p>",
  "messages": [
    {
      "id": 2895915,
      "postDate": "2024-06-29T12:56:14.457Z",
      "content": "<p>I’ve tried masking the targets and it wasn’t helpful for me.</p>",
      "rawMarkdown": "I’ve tried masking the targets and it wasn’t helpful for me.",
      "votes": 2
    },
    {
      "id": 2895496,
      "postDate": "2024-06-29T08:13:40.450Z",
      "content": "<p>Hello guys, i have my main model and i decided to do an experiment and fork it to predict only for level 41 (one that i have the worst result). I let it train whole night and i was pretty sure i would see a small improvement in the loss of variables of level 41 but i did not. That was surprising for me, i also tried to predict all levels of one specific variable (this time i let it run only for a few iterations) and it did not seem promising. Anybody that tried this and got better result? Does it make sense that trying to predict much less targets did not produced better results?</p>",
      "rawMarkdown": "Hello guys, i have my main model and i decided to do an experiment and fork it to predict only for level 41 (one that i have the worst result). I let it train whole night and i was pretty sure i would see a small improvement in the loss of variables of level 41 but i did not. That was surprising for me, i also tried to predict all levels of one specific variable (this time i let it run only for a few iterations) and it did not seem promising. Anybody that tried this and got better result? Does it make sense that trying to predict much less targets did not produced better results?",
      "votes": 2
    }
  ],
  "comments": [
    {
      "id": 2895915,
      "author_name": "Rob Freeman",
      "author_url": "",
      "post_date": "2024-06-29T12:56:14.457000",
      "content": "<p>I’ve tried masking the targets and it wasn’t helpful for me.</p>",
      "votes": 2,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2895915": "I’ve tried masking the targets and it wasn’t helpful for me.",
    "2895496": "Hello guys, i have my main model and i decided to do an experiment and fork it to predict only for level 41 (one that i have the worst result). I let it train whole night and i was pretty sure i would see a small improvement in the loss of variables of level 41 but i did not. That was surprising for me, i also tried to predict all levels of one specific variable (this time i let it run only for a few iterations) and it did not seem promising. Anybody that tried this and got better result? Does it make sense that trying to predict much less targets did not produced better results?"
  }
}