{
  "id": 663021,
  "title": "Looking forward to see the top solutions",
  "url": "/competitions/MABe-mouse-behavior-detection/discussion/663021",
  "author_name": "",
  "post_date": "2025-12-15T23:49:53.321480500Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>My base models are quite weak. The NN model scores around 0.42 and the tree model 0.40. I threw some optimization + post-processing together today and got it to 0.495 the last minute. I am wondering if the trick is stronger base models or better optimization techniques or both :)</p>",
  "messages": [
    {
      "id": "3377206",
      "postDate": "12/15/2025 23:49:53",
      "content": "<p>My base models are quite weak. The NN model scores around 0.42 and the tree model 0.40. I threw some optimization + post-processing together today and got it to 0.495 the last minute. I am wondering if the trick is stronger base models or better optimization techniques or both :)</p>",
      "rawMarkdown": "My base models are quite weak. The NN model scores around 0.42 and the tree model 0.40. I threw some optimization + post-processing together today and got it to 0.495 the last minute. I am wondering if the trick is stronger base models or better optimization techniques or both :)",
      "votes": null
    },
    {
      "id": "3377213",
      "postDate": "12/15/2025 23:57:13",
      "content": "<p>I would be surprised if all of the top 10 contestants aren't utilizing some form of deep learning models. My best single NN model  scores ~ 0.542 on public LB. The winner most likely found something else. My guess is that there is something particular about the dataset what others missed and/or found a way to get something out of semi-supervised learning.</p>",
      "rawMarkdown": "I would be surprised if all of the top 10 contestants aren't utilizing some form of deep learning models. My best single NN model  scores ~ 0.542 on public LB. The winner most likely found something else. My guess is that there is something particular about the dataset what others missed and/or found a way to get something out of semi-supervised learning.",
      "votes": null
    },
    {
      "id": "3377217",
      "postDate": "12/15/2025 23:59:20",
      "content": "<p>Nice. That's super strong. Both of my models absolutely struggled with certain labs/rare actions. Looking forward to see how people managed to deal with this.</p>",
      "rawMarkdown": "Nice. That's super strong. Both of my models absolutely struggled with certain labs/rare actions. Looking forward to see how people managed to deal with this.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3377213,
      "author_name": "herrahuu",
      "author_url": "",
      "post_date": "12/15/2025 23:57:13",
      "content": "<p>I would be surprised if all of the top 10 contestants aren't utilizing some form of deep learning models. My best single NN model  scores ~ 0.542 on public LB. The winner most likely found something else. My guess is that there is something particular about the dataset what others missed and/or found a way to get something out of semi-supervised learning.</p>",
      "votes": null,
      "replies": [
        {
          "id": 3377217,
          "author_name": "xiaoz259",
          "author_url": "",
          "post_date": "12/15/2025 23:59:20",
          "content": "<p>Nice. That's super strong. Both of my models absolutely struggled with certain labs/rare actions. Looking forward to see how people managed to deal with this.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3377206": "My base models are quite weak. The NN model scores around 0.42 and the tree model 0.40. I threw some optimization + post-processing together today and got it to 0.495 the last minute. I am wondering if the trick is stronger base models or better optimization techniques or both :)",
    "3377213": "I would be surprised if all of the top 10 contestants aren't utilizing some form of deep learning models. My best single NN model  scores ~ 0.542 on public LB. The winner most likely found something else. My guess is that there is something particular about the dataset what others missed and/or found a way to get something out of semi-supervised learning.",
    "3377217": "Nice. That's super strong. Both of my models absolutely struggled with certain labs/rare actions. Looking forward to see how people managed to deal with this."
  },
  "source": "meta"
}