{
  "id": 663049,
  "title": "27th place solution",
  "url": "/competitions/MABe-mouse-behavior-detection/writeups/27th-place-solution",
  "author_name": "",
  "post_date": "2025-12-16T03:53:44.987Z",
  "votes": 18,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I started great notebook. Thank you <a href=\"https://www.kaggle.com/ravaghi\" target=\"_blank\">@ravaghi</a>.\n<a href=\"https://www.kaggle.com/code/ravaghi/social-action-recognition-in-mice-xgboost\" target=\"_blank\">https://www.kaggle.com/code/ravaghi/social-action-recognition-in-mice-xgboost</a></p>",
  "messages": [
    {
      "id": "3377306",
      "postDate": "12/16/2025 03:52:53",
      "content": "<p>I started great notebook. Thank you <a href=\"https://www.kaggle.com/ravaghi\" target=\"_blank\">@ravaghi</a>.\n<a href=\"https://www.kaggle.com/code/ravaghi/social-action-recognition-in-mice-xgboost\" target=\"_blank\">https://www.kaggle.com/code/ravaghi/social-action-recognition-in-mice-xgboost</a></p>",
      "rawMarkdown": "I started great notebook. Thank you @ravaghi.\nhttps://www.kaggle.com/code/ravaghi/social-action-recognition-in-mice-xgboost",
      "votes": null
    },
    {
      "id": "3377378",
      "postDate": "12/16/2025 07:32:20",
      "content": "<p>\"Congrats shinmura, amazing solution! I have a couple of questions about the details:</p>\n<p>1.For Improvement 3 (per-Dataset Training), how did you separate or define the different datasets?\n2.In Improvement 4, how well did the nfnet model perform? </p>",
      "rawMarkdown": "\"Congrats shinmura, amazing solution! I have a couple of questions about the details:\n\n1.For Improvement 3 (per-Dataset Training), how did you separate or define the different datasets?\n2.In Improvement 4, how well did the nfnet model perform?",
      "votes": null
    },
    {
      "id": "3377855",
      "postDate": "12/17/2025 03:58:33",
      "content": "<ol>\n<li><p>Separate datasets (e.g., Adaptable Snail, Boisterous Parrot).</p></li>\n<li><p>nfnet model CV was around 0.2. But adding stacking this output improved accuracy for some datasets. This was particularly effective for datasets with small amounts of data. Since the image model was trained across datasets (using a large amount of data), it was likely effective for datasets with small amounts of data.</p></li>\n</ol>",
      "rawMarkdown": "1. Separate datasets (e.g., Adaptable Snail, Boisterous Parrot).\n\n2. nfnet model CV was around 0.2. But adding stacking this output improved accuracy for some datasets. This was particularly effective for datasets with small amounts of data. Since the image model was trained across datasets (using a large amount of data), it was likely effective for datasets with small amounts of data.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3377378,
      "author_name": "ybzjdbc",
      "author_url": "",
      "post_date": "12/16/2025 07:32:20",
      "content": "<p>\"Congrats shinmura, amazing solution! I have a couple of questions about the details:</p>\n<p>1.For Improvement 3 (per-Dataset Training), how did you separate or define the different datasets?\n2.In Improvement 4, how well did the nfnet model perform? </p>",
      "votes": null,
      "replies": [
        {
          "id": 3377855,
          "author_name": "shinmurashinmura",
          "author_url": "",
          "post_date": "12/17/2025 03:58:33",
          "content": "<ol>\n<li><p>Separate datasets (e.g., Adaptable Snail, Boisterous Parrot).</p></li>\n<li><p>nfnet model CV was around 0.2. But adding stacking this output improved accuracy for some datasets. This was particularly effective for datasets with small amounts of data. Since the image model was trained across datasets (using a large amount of data), it was likely effective for datasets with small amounts of data.</p></li>\n</ol>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3377306": "I started great notebook. Thank you @ravaghi.\nhttps://www.kaggle.com/code/ravaghi/social-action-recognition-in-mice-xgboost",
    "3377378": "\"Congrats shinmura, amazing solution! I have a couple of questions about the details:\n\n1.For Improvement 3 (per-Dataset Training), how did you separate or define the different datasets?\n2.In Improvement 4, how well did the nfnet model perform?",
    "3377855": "1. Separate datasets (e.g., Adaptable Snail, Boisterous Parrot).\n\n2. nfnet model CV was around 0.2. But adding stacking this output improved accuracy for some datasets. This was particularly effective for datasets with small amounts of data. Since the image model was trained across datasets (using a large amount of data), it was likely effective for datasets with small amounts of data."
  },
  "source": "meta"
}