{
  "id": 533301,
  "title": "Which sigma_true should we use?",
  "url": "/competitions/ariel-data-challenge-2024/discussion/533301",
  "author_name": "Johnny Lee",
  "post_date": "2024-09-10T13:30:37.101000",
  "votes": 8,
  "comment_count": 2,
  "views": 0,
  "content": "<p>We can find the full metric implementation <a href=\"https://www.kaggle.com/code/metric/ariel-gaussian-log-likelihood\" target=\"_blank\">here</a>, but there is no information about the sigma_true. Almost all the public notebooks are using 0.000003. In my case it is a little lower than public LB. Maybe the host should provide the value of sigma_true. <a href=\"https://www.kaggle.com/gordonyip\" target=\"_blank\">@gordonyip</a> </p>",
  "messages": [
    {
      "id": 2985272,
      "postDate": "2024-09-10T13:41:10.460Z",
      "content": "<p>From the competition description<br>\n\"We define L_ideal as the case where the submission perfectly matches the ground truth values, with an uncertainty of 10 parts per million (ppm). \"<br>\n10ppm = 1e-5<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F349155%2F560aafbca6e954284cf9c2b0a8083b20%2FGLL_Ideal.PNG?generation=1725975667413355&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "From the competition description\n\"We define L_ideal as the case where the submission perfectly matches the ground truth values, with an uncertainty of 10 parts per million (ppm). \"\n10ppm = 1e-5\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F349155%2F560aafbca6e954284cf9c2b0a8083b20%2FGLL_Ideal.PNG?generation=1725975667413355&alt=media)",
      "votes": 7,
      "replies": [
        {
          "id": 2985276,
          "postDate": "2024-09-10T13:44:06.490Z",
          "content": "<p>Thank you. 😀</p>",
          "rawMarkdown": "Thank you. 😀"
        }
      ]
    },
    {
      "id": 2985256,
      "postDate": "2024-09-10T13:30:37.103Z",
      "content": "<p>We can find the full metric implementation <a href=\"https://www.kaggle.com/code/metric/ariel-gaussian-log-likelihood\" target=\"_blank\">here</a>, but there is no information about the sigma_true. Almost all the public notebooks are using 0.000003. In my case it is a little lower than public LB. Maybe the host should provide the value of sigma_true. <a href=\"https://www.kaggle.com/gordonyip\" target=\"_blank\">@gordonyip</a> </p>",
      "rawMarkdown": "We can find the full metric implementation [here](https://www.kaggle.com/code/metric/ariel-gaussian-log-likelihood), but there is no information about the sigma_true. Almost all the public notebooks are using 0.000003. In my case it is a little lower than public LB. Maybe the host should provide the value of sigma_true. @gordonyip \n",
      "votes": 8
    }
  ],
  "comments": [
    {
      "id": 2985272,
      "author_name": "DennisSakva",
      "author_url": "",
      "post_date": "2024-09-10T13:41:10.460000",
      "content": "<p>From the competition description<br>\n\"We define L_ideal as the case where the submission perfectly matches the ground truth values, with an uncertainty of 10 parts per million (ppm). \"<br>\n10ppm = 1e-5<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F349155%2F560aafbca6e954284cf9c2b0a8083b20%2FGLL_Ideal.PNG?generation=1725975667413355&amp;alt=media\" alt=\"\"></p>",
      "votes": 7,
      "replies": [
        {
          "id": 2985276,
          "author_name": "Johnny Lee",
          "author_url": "",
          "post_date": "2024-09-10T13:44:06.490000",
          "content": "<p>Thank you. 😀</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2985272": "From the competition description\n\"We define L_ideal as the case where the submission perfectly matches the ground truth values, with an uncertainty of 10 parts per million (ppm). \"\n10ppm = 1e-5\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F349155%2F560aafbca6e954284cf9c2b0a8083b20%2FGLL_Ideal.PNG?generation=1725975667413355&alt=media)",
    "2985256": "We can find the full metric implementation [here](https://www.kaggle.com/code/metric/ariel-gaussian-log-likelihood), but there is no information about the sigma_true. Almost all the public notebooks are using 0.000003. In my case it is a little lower than public LB. Maybe the host should provide the value of sigma_true. @gordonyip \n"
  }
}