{
  "id": 554101,
  "title": "Online learning?",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/554101",
  "author_name": "Nicholas Lyu",
  "post_date": "2024-12-30T12:34:07.436000",
  "votes": 6,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Curious about how much top teams have benefited from online learning, since there will be a tradeoff between ensembling more (static) models vs. having fewer adaptive models. </p>\n<p>Below is the profile of an oof model (nn) validated on partition_9, with y-axis being the weighted r2 of the trades on that day. The model has lower r2 after 60 days into the future, though I'm not sure if this is due to consistently evolving market conditions or just variance. How much has online learning added to your CV / LB?</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1943421%2Feab549bbd5f383fc288652f8e4981a69%2FScreenshot%202024-12-30%20at%208.32.06PM.png?generation=1735561938443271&amp;alt=media\" alt=\"\"></p>",
  "messages": [
    {
      "id": 3084132,
      "postDate": "2024-12-30T12:34:07.437Z",
      "content": "<p>Curious about how much top teams have benefited from online learning, since there will be a tradeoff between ensembling more (static) models vs. having fewer adaptive models. </p>\n<p>Below is the profile of an oof model (nn) validated on partition_9, with y-axis being the weighted r2 of the trades on that day. The model has lower r2 after 60 days into the future, though I'm not sure if this is due to consistently evolving market conditions or just variance. How much has online learning added to your CV / LB?</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1943421%2Feab549bbd5f383fc288652f8e4981a69%2FScreenshot%202024-12-30%20at%208.32.06PM.png?generation=1735561938443271&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "Curious about how much top teams have benefited from online learning, since there will be a tradeoff between ensembling more (static) models vs. having fewer adaptive models. \n\nBelow is the profile of an oof model (nn) validated on partition_9, with y-axis being the weighted r2 of the trades on that day. The model has lower r2 after 60 days into the future, though I'm not sure if this is due to consistently evolving market conditions or just variance. How much has online learning added to your CV / LB?\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1943421%2Feab549bbd5f383fc288652f8e4981a69%2FScreenshot%202024-12-30%20at%208.32.06PM.png?generation=1735561938443271&alt=media)",
      "votes": 6
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3084132": "Curious about how much top teams have benefited from online learning, since there will be a tradeoff between ensembling more (static) models vs. having fewer adaptive models. \n\nBelow is the profile of an oof model (nn) validated on partition_9, with y-axis being the weighted r2 of the trades on that day. The model has lower r2 after 60 days into the future, though I'm not sure if this is due to consistently evolving market conditions or just variance. How much has online learning added to your CV / LB?\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1943421%2Feab549bbd5f383fc288652f8e4981a69%2FScreenshot%202024-12-30%20at%208.32.06PM.png?generation=1735561938443271&alt=media)"
  }
}