{
  "id": 454177,
  "title": "Larger bounds = faster?",
  "url": "/competitions/predict-ai-model-runtime/discussion/454177",
  "author_name": "Louka Ewington-Pitsos",
  "post_date": "2023-11-09T09:28:10.553000",
  "votes": 3,
  "comment_count": 0,
  "views": 0,
  "content": "<p>It seems like for the tiling dataset, runtime is significantly negatively correlated with many of the tiling features, such as <code>output_bounds_sum</code> as shown below over all graphs in the train collection. </p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1577010%2F6d16832c244645fc5efcbf3d7046d26c%2FScreenshot%202023-11-09%20202451.png?generation=1699521911964360&amp;alt=media\" alt=\"\"></p>\n<p><a href=\"https://www.kaggle.com/code/lewington/larger-tile-bounds-faster\" target=\"_blank\">https://www.kaggle.com/code/lewington/larger-tile-bounds-faster</a></p>\n<p>I also used linear regression over 4 of these features to make a better-than-random submission <a href=\"https://www.kaggle.com/code/lewington/linear-regression-baseline\" target=\"_blank\">https://www.kaggle.com/code/lewington/linear-regression-baseline</a> which is a bit insane</p>\n<p>I went over both notebooks in this video: <a href=\"https://youtu.be/53P7u6IDZnc\" target=\"_blank\">https://youtu.be/53P7u6IDZnc</a> </p>\n<p>If I'm missing something obvious please let me know</p>",
  "messages": [
    {
      "id": 2518414,
      "postDate": "2023-11-09T09:28:10.553Z",
      "content": "<p>It seems like for the tiling dataset, runtime is significantly negatively correlated with many of the tiling features, such as <code>output_bounds_sum</code> as shown below over all graphs in the train collection. </p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1577010%2F6d16832c244645fc5efcbf3d7046d26c%2FScreenshot%202023-11-09%20202451.png?generation=1699521911964360&amp;alt=media\" alt=\"\"></p>\n<p><a href=\"https://www.kaggle.com/code/lewington/larger-tile-bounds-faster\" target=\"_blank\">https://www.kaggle.com/code/lewington/larger-tile-bounds-faster</a></p>\n<p>I also used linear regression over 4 of these features to make a better-than-random submission <a href=\"https://www.kaggle.com/code/lewington/linear-regression-baseline\" target=\"_blank\">https://www.kaggle.com/code/lewington/linear-regression-baseline</a> which is a bit insane</p>\n<p>I went over both notebooks in this video: <a href=\"https://youtu.be/53P7u6IDZnc\" target=\"_blank\">https://youtu.be/53P7u6IDZnc</a> </p>\n<p>If I'm missing something obvious please let me know</p>",
      "rawMarkdown": "It seems like for the tiling dataset, runtime is significantly negatively correlated with many of the tiling features, such as `output_bounds_sum` as shown below over all graphs in the train collection. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1577010%2F6d16832c244645fc5efcbf3d7046d26c%2FScreenshot%202023-11-09%20202451.png?generation=1699521911964360&alt=media)\n\n\nhttps://www.kaggle.com/code/lewington/larger-tile-bounds-faster\n\nI also used linear regression over 4 of these features to make a better-than-random submission https://www.kaggle.com/code/lewington/linear-regression-baseline which is a bit insane\n\nI went over both notebooks in this video: https://youtu.be/53P7u6IDZnc \n\nIf I'm missing something obvious please let me know",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2518414": "It seems like for the tiling dataset, runtime is significantly negatively correlated with many of the tiling features, such as `output_bounds_sum` as shown below over all graphs in the train collection. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1577010%2F6d16832c244645fc5efcbf3d7046d26c%2FScreenshot%202023-11-09%20202451.png?generation=1699521911964360&alt=media)\n\n\nhttps://www.kaggle.com/code/lewington/larger-tile-bounds-faster\n\nI also used linear regression over 4 of these features to make a better-than-random submission https://www.kaggle.com/code/lewington/linear-regression-baseline which is a bit insane\n\nI went over both notebooks in this video: https://youtu.be/53P7u6IDZnc \n\nIf I'm missing something obvious please let me know"
  }
}