{
  "id": 543054,
  "title": "Clarifying Submission of Predictions and Uncertainties for GLL Scoring",
  "url": "/competitions/ariel-data-challenge-2024/discussion/543054",
  "author_name": "Cyrus",
  "post_date": "2024-10-28T12:07:55.591000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>The project involves predicting values for 283 data points and submitting these predictions along with corresponding uncertainties (sigma). These are used to compute the Generalized Log-Likelihood (GLL) score. However, the objective of the project and how to properly submit the results (pred and sigma) are not entirely clear to me.<br>\nI've noticed that others estimate the mean values and use a small sigma with some adjustments. If we can potentially estimate predictions for the 283 values using machine learning (ML) or deep learning (DL) methods, should we consider the mean of these predictions as 'pred' and the variation (e.g., standard deviation) as 'sigma'? How should we generate and submit 'pred' and 'sigma' in this context, and how are they used in calculating the GLL score?</p>",
  "messages": [
    {
      "id": 3030307,
      "postDate": "2024-10-28T12:07:55.590Z",
      "content": "<p>The project involves predicting values for 283 data points and submitting these predictions along with corresponding uncertainties (sigma). These are used to compute the Generalized Log-Likelihood (GLL) score. However, the objective of the project and how to properly submit the results (pred and sigma) are not entirely clear to me.<br>\nI've noticed that others estimate the mean values and use a small sigma with some adjustments. If we can potentially estimate predictions for the 283 values using machine learning (ML) or deep learning (DL) methods, should we consider the mean of these predictions as 'pred' and the variation (e.g., standard deviation) as 'sigma'? How should we generate and submit 'pred' and 'sigma' in this context, and how are they used in calculating the GLL score?</p>",
      "rawMarkdown": "The project involves predicting values for 283 data points and submitting these predictions along with corresponding uncertainties (sigma). These are used to compute the Generalized Log-Likelihood (GLL) score. However, the objective of the project and how to properly submit the results (pred and sigma) are not entirely clear to me.\nI've noticed that others estimate the mean values and use a small sigma with some adjustments. If we can potentially estimate predictions for the 283 values using machine learning (ML) or deep learning (DL) methods, should we consider the mean of these predictions as 'pred' and the variation (e.g., standard deviation) as 'sigma'? How should we generate and submit 'pred' and 'sigma' in this context, and how are they used in calculating the GLL score?"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3030307": "The project involves predicting values for 283 data points and submitting these predictions along with corresponding uncertainties (sigma). These are used to compute the Generalized Log-Likelihood (GLL) score. However, the objective of the project and how to properly submit the results (pred and sigma) are not entirely clear to me.\nI've noticed that others estimate the mean values and use a small sigma with some adjustments. If we can potentially estimate predictions for the 283 values using machine learning (ML) or deep learning (DL) methods, should we consider the mean of these predictions as 'pred' and the variation (e.g., standard deviation) as 'sigma'? How should we generate and submit 'pred' and 'sigma' in this context, and how are they used in calculating the GLL score?"
  }
}