{
  "id": 567374,
  "title": "First week RNA 3D folding Learning +  Second week plan",
  "url": "/competitions/stanford-rna-3d-folding/discussion/567374",
  "author_name": "Pastor Soto",
  "post_date": "2025-03-10T00:25:40.594000",
  "votes": 3,
  "comment_count": 0,
  "views": 0,
  "content": "<p>The competition is harder than expected. I am trying to understand some code that was used to train a basic pipeline with feature engineering, but things are more challenging than I anticipated. Additionally, making submissions is not as easy as I had planned.</p>\n<h1>First Week Learning</h1>\n<ol>\n<li><p><strong>RNA 3D Folding</strong>: This involves using an RNA sequence to predict how it folds in XYZ space.</p></li>\n<li><p><strong>Accuracy of Human Experts</strong>: Based on the leaderboard, the accuracy of human experts is around 0.49 in Template Modeling, which ranges from 0 to 1 (higher is better). My first attempt achieved a score of 0.027 using a copy of this <a href=\"https://www.kaggle.com/code/olaflundstrom/stanford-rna-3d-folding-kaggle-competition/notebook\" target=\"_blank\">notebook</a>.</p></li>\n<li><p><strong>Computational Power</strong>: Discussions suggest that winning might require significant computational power. However, I am not overly concerned about this, as my goal is to have a decent competition experience while sharing my learning journey.</p></li>\n<li><p><strong>Current Approach</strong>: My current strategy is to take the mentioned notebook, go through the code line by line, and try to understand it using LLMs to replicate in my notebook (I am taking learning suggestions since I don't think might be the most optimal way of doing it)</p></li>\n</ol>\n<h1>Second Week Plan</h1>\n<ul>\n<li>Finish analyzing the first notebook and make one more submission.  </li>\n<li>Create a simple solution on my own, now that I understand the basics of the competition, modeling, and data.  </li>\n<li>Evaluate different learning approaches that can be used in this competition.  </li>\n<li>Engage in discussions with other participants and try to incorporate learning approaches from their posts.  </li>\n</ul>",
  "messages": [
    {
      "id": 3145537,
      "postDate": "2025-03-10T00:25:40.593Z",
      "content": "<p>The competition is harder than expected. I am trying to understand some code that was used to train a basic pipeline with feature engineering, but things are more challenging than I anticipated. Additionally, making submissions is not as easy as I had planned.</p>\n<h1>First Week Learning</h1>\n<ol>\n<li><p><strong>RNA 3D Folding</strong>: This involves using an RNA sequence to predict how it folds in XYZ space.</p></li>\n<li><p><strong>Accuracy of Human Experts</strong>: Based on the leaderboard, the accuracy of human experts is around 0.49 in Template Modeling, which ranges from 0 to 1 (higher is better). My first attempt achieved a score of 0.027 using a copy of this <a href=\"https://www.kaggle.com/code/olaflundstrom/stanford-rna-3d-folding-kaggle-competition/notebook\" target=\"_blank\">notebook</a>.</p></li>\n<li><p><strong>Computational Power</strong>: Discussions suggest that winning might require significant computational power. However, I am not overly concerned about this, as my goal is to have a decent competition experience while sharing my learning journey.</p></li>\n<li><p><strong>Current Approach</strong>: My current strategy is to take the mentioned notebook, go through the code line by line, and try to understand it using LLMs to replicate in my notebook (I am taking learning suggestions since I don't think might be the most optimal way of doing it)</p></li>\n</ol>\n<h1>Second Week Plan</h1>\n<ul>\n<li>Finish analyzing the first notebook and make one more submission.  </li>\n<li>Create a simple solution on my own, now that I understand the basics of the competition, modeling, and data.  </li>\n<li>Evaluate different learning approaches that can be used in this competition.  </li>\n<li>Engage in discussions with other participants and try to incorporate learning approaches from their posts.  </li>\n</ul>",
      "rawMarkdown": "The competition is harder than expected. I am trying to understand some code that was used to train a basic pipeline with feature engineering, but things are more challenging than I anticipated. Additionally, making submissions is not as easy as I had planned.\n\n# First Week Learning\n\n1. **RNA 3D Folding**: This involves using an RNA sequence to predict how it folds in XYZ space.\n\n2. **Accuracy of Human Experts**: Based on the leaderboard, the accuracy of human experts is around 0.49 in Template Modeling, which ranges from 0 to 1 (higher is better). My first attempt achieved a score of 0.027 using a copy of this [notebook](https://www.kaggle.com/code/olaflundstrom/stanford-rna-3d-folding-kaggle-competition/notebook).\n\n3. **Computational Power**: Discussions suggest that winning might require significant computational power. However, I am not overly concerned about this, as my goal is to have a decent competition experience while sharing my learning journey.\n\n4. **Current Approach**: My current strategy is to take the mentioned notebook, go through the code line by line, and try to understand it using LLMs to replicate in my notebook (I am taking learning suggestions since I don't think might be the most optimal way of doing it)\n\n# Second Week Plan\n\n- Finish analyzing the first notebook and make one more submission.  \n- Create a simple solution on my own, now that I understand the basics of the competition, modeling, and data.  \n- Evaluate different learning approaches that can be used in this competition.  \n- Engage in discussions with other participants and try to incorporate learning approaches from their posts.  ",
      "votes": 3
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3145537": "The competition is harder than expected. I am trying to understand some code that was used to train a basic pipeline with feature engineering, but things are more challenging than I anticipated. Additionally, making submissions is not as easy as I had planned.\n\n# First Week Learning\n\n1. **RNA 3D Folding**: This involves using an RNA sequence to predict how it folds in XYZ space.\n\n2. **Accuracy of Human Experts**: Based on the leaderboard, the accuracy of human experts is around 0.49 in Template Modeling, which ranges from 0 to 1 (higher is better). My first attempt achieved a score of 0.027 using a copy of this [notebook](https://www.kaggle.com/code/olaflundstrom/stanford-rna-3d-folding-kaggle-competition/notebook).\n\n3. **Computational Power**: Discussions suggest that winning might require significant computational power. However, I am not overly concerned about this, as my goal is to have a decent competition experience while sharing my learning journey.\n\n4. **Current Approach**: My current strategy is to take the mentioned notebook, go through the code line by line, and try to understand it using LLMs to replicate in my notebook (I am taking learning suggestions since I don't think might be the most optimal way of doing it)\n\n# Second Week Plan\n\n- Finish analyzing the first notebook and make one more submission.  \n- Create a simple solution on my own, now that I understand the basics of the competition, modeling, and data.  \n- Evaluate different learning approaches that can be used in this competition.  \n- Engage in discussions with other participants and try to incorporate learning approaches from their posts.  "
  }
}