{
  "id": 603217,
  "title": "How does h-blend select weights?",
  "url": "/competitions/cmi-detect-behavior-with-sensor-data/discussion/603217",
  "author_name": "",
  "post_date": "2025-09-01T08:39:57.814140900Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I am stuck using an ensemble model like h-blend. When using average weight, I can tell the difference and how to set the weight, but with h-blend, I totally have no idea. Changing models won't get a higher score, changing weight won't get a higher score, even keeping everything the same, just changing the weight will make the score lower. And everyone in the public notebook is using the same weight! Just wondering why.</p>",
  "messages": [
    {
      "id": "3279699",
      "postDate": "09/01/2025 08:39:57",
      "content": "<p>I am stuck using an ensemble model like h-blend. When using average weight, I can tell the difference and how to set the weight, but with h-blend, I totally have no idea. Changing models won't get a higher score, changing weight won't get a higher score, even keeping everything the same, just changing the weight will make the score lower. And everyone in the public notebook is using the same weight! Just wondering why.</p>",
      "rawMarkdown": "I am stuck using an ensemble model like h-blend. When using average weight, I can tell the difference and how to set the weight, but with h-blend, I totally have no idea. Changing models won't get a higher score, changing weight won't get a higher score, even keeping everything the same, just changing the weight will make the score lower. And everyone in the public notebook is using the same weight! Just wondering why.",
      "votes": null
    },
    {
      "id": "3279726",
      "postDate": "09/01/2025 10:24:57",
      "content": "<p>This is not a scientific method and may even be avoided for correct model practice <a href=\"https://www.kaggle.com/wliilamsam\" target=\"_blank\">@wliilamsam</a> </p>",
      "rawMarkdown": "This is not a scientific method and may even be avoided for correct model practice @wliilamsam",
      "votes": null
    },
    {
      "id": "3280622",
      "postDate": "09/03/2025 00:29:50",
      "content": "<p>Thank u for ur comment that makes me believe myself and got shakedup</p>",
      "rawMarkdown": "Thank u for ur comment that makes me believe myself and got shakedup",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3279726,
      "author_name": "ravi20076",
      "author_url": "",
      "post_date": "09/01/2025 10:24:57",
      "content": "<p>This is not a scientific method and may even be avoided for correct model practice <a href=\"https://www.kaggle.com/wliilamsam\" target=\"_blank\">@wliilamsam</a> </p>",
      "votes": null,
      "replies": [
        {
          "id": 3280622,
          "author_name": "wliilamsam",
          "author_url": "",
          "post_date": "09/03/2025 00:29:50",
          "content": "<p>Thank u for ur comment that makes me believe myself and got shakedup</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3279699": "I am stuck using an ensemble model like h-blend. When using average weight, I can tell the difference and how to set the weight, but with h-blend, I totally have no idea. Changing models won't get a higher score, changing weight won't get a higher score, even keeping everything the same, just changing the weight will make the score lower. And everyone in the public notebook is using the same weight! Just wondering why.",
    "3279726": "This is not a scientific method and may even be avoided for correct model practice @wliilamsam",
    "3280622": "Thank u for ur comment that makes me believe myself and got shakedup"
  },
  "source": "meta"
}