{
  "id": 549871,
  "title": "Is it overfitting? Modifying any parameters will lead to a decrease in performance",
  "url": "/competitions/child-mind-institute-problematic-internet-use/discussion/549871",
  "author_name": "Less",
  "post_date": "2024-12-04T09:10:59.113000",
  "votes": 5,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Is it overfitting? Modifying any parameters will lead to a decrease in performance. What else can we do about open code？</p>",
  "messages": [
    {
      "id": 3063216,
      "postDate": "2024-12-04T09:10:59.113Z",
      "content": "<p>Is it overfitting? Modifying any parameters will lead to a decrease in performance. What else can we do about open code？</p>",
      "rawMarkdown": "Is it overfitting? Modifying any parameters will lead to a decrease in performance. What else can we do about open code？",
      "votes": 5
    },
    {
      "id": 3063271,
      "postDate": "2024-12-04T10:26:38.753Z",
      "content": "<p>Simple <a href=\"https://www.kaggle.com/jiaoyouzhang\" target=\"_blank\">@jiaoyouzhang</a> <br>\nDon't use it!</p>",
      "rawMarkdown": "Simple @jiaoyouzhang \nDon't use it!",
      "votes": 6
    },
    {
      "id": 3064646,
      "postDate": "2024-12-05T20:28:26.160Z",
      "content": "<p>I think there are some outliers, possibly bad or misleading data and so results are not very stable. Even if you look at the results from different folds in the CV, you can see that they vary a lot. I am not sure the best way to deal with outliers, or even to spot them efficiently.</p>",
      "rawMarkdown": "I think there are some outliers, possibly bad or misleading data and so results are not very stable. Even if you look at the results from different folds in the CV, you can see that they vary a lot. I am not sure the best way to deal with outliers, or even to spot them efficiently.",
      "votes": 1
    },
    {
      "id": 3076485,
      "postDate": "2024-12-20T01:55:24.993Z",
      "rawMarkdown": "",
      "isDeleted": true
    },
    {
      "id": 3064764,
      "postDate": "2024-12-06T01:14:00.190Z",
      "rawMarkdown": "",
      "isDeleted": true,
      "replies": [
        {
          "id": 3065247,
          "postDate": "2024-12-06T15:21:29.253Z",
          "content": "<p>If changing the parameters leads to decreased performance, it may suggest that the model is overly customized to the specific patterns of the training data instead of learning generalizable features. An overly customized model is less flexible and may struggle to adapt to new scenarios or datasets. You may<br>\nApply training in batches.<br>\nMonitor the validation loss parallelly.</p>",
          "rawMarkdown": "If changing the parameters leads to decreased performance, it may suggest that the model is overly customized to the specific patterns of the training data instead of learning generalizable features. An overly customized model is less flexible and may struggle to adapt to new scenarios or datasets. You may\nApply training in batches.\nMonitor the validation loss parallelly.\n",
          "votes": 1
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 3063271,
      "author_name": "Ravi Ramakrishnan",
      "author_url": "",
      "post_date": "2024-12-04T10:26:38.753000",
      "content": "<p>Simple <a href=\"https://www.kaggle.com/jiaoyouzhang\" target=\"_blank\">@jiaoyouzhang</a> <br>\nDon't use it!</p>",
      "votes": 6,
      "replies": []
    },
    {
      "id": 3064646,
      "author_name": "Lawrence Chernin",
      "author_url": "",
      "post_date": "2024-12-05T20:28:26.160000",
      "content": "<p>I think there are some outliers, possibly bad or misleading data and so results are not very stable. Even if you look at the results from different folds in the CV, you can see that they vary a lot. I am not sure the best way to deal with outliers, or even to spot them efficiently.</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 3076485,
      "author_name": "",
      "author_url": "",
      "post_date": "2024-12-20T01:55:24.993000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 3064764,
      "author_name": "",
      "author_url": "",
      "post_date": "2024-12-06T01:14:00.190000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 3065247,
          "author_name": "Yash Mishra",
          "author_url": "",
          "post_date": "2024-12-06T15:21:29.253000",
          "content": "<p>If changing the parameters leads to decreased performance, it may suggest that the model is overly customized to the specific patterns of the training data instead of learning generalizable features. An overly customized model is less flexible and may struggle to adapt to new scenarios or datasets. You may<br>\nApply training in batches.<br>\nMonitor the validation loss parallelly.</p>",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3063216": "Is it overfitting? Modifying any parameters will lead to a decrease in performance. What else can we do about open code？",
    "3063271": "Simple @jiaoyouzhang \nDon't use it!",
    "3064646": "I think there are some outliers, possibly bad or misleading data and so results are not very stable. Even if you look at the results from different folds in the CV, you can see that they vary a lot. I am not sure the best way to deal with outliers, or even to spot them efficiently.",
    "3076485": "",
    "3064764": ""
  }
}