{
  "id": 447025,
  "title": "perfect prediction of \"is_tma\"",
  "url": "/competitions/UBC-OCEAN/discussion/447025",
  "author_name": "ISAKA Tsuyoshi",
  "post_date": "2023-10-14T02:11:37.381000",
  "votes": 10,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Using a threshold value of around 5000 for both <code>image_width</code> and <code>image_height</code> seems to perfectly predict the value of <code>is_tma</code>.</p>\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/2481282/19785/2023-10-14%20110601.png\" alt=\"Scatter Plot of image_width and image_height colored by is_tma\"></p>\n<p>For more details, please refer to the notebook below.</p>\n<p><a href=\"https://www.kaggle.com/isakatsuyoshi/perfect-prediction-of-is-tma\" target=\"_blank\">https://www.kaggle.com/isakatsuyoshi/perfect-prediction-of-is-tma</a></p>",
  "messages": [
    {
      "id": 2481282,
      "postDate": "2023-10-14T02:11:37.380Z",
      "content": "<p>Using a threshold value of around 5000 for both <code>image_width</code> and <code>image_height</code> seems to perfectly predict the value of <code>is_tma</code>.</p>\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/2481282/19785/2023-10-14%20110601.png\" alt=\"Scatter Plot of image_width and image_height colored by is_tma\"></p>\n<p>For more details, please refer to the notebook below.</p>\n<p><a href=\"https://www.kaggle.com/isakatsuyoshi/perfect-prediction-of-is-tma\" target=\"_blank\">https://www.kaggle.com/isakatsuyoshi/perfect-prediction-of-is-tma</a></p>",
      "rawMarkdown": "Using a threshold value of around 5000 for both `image_width` and `image_height` seems to perfectly predict the value of `is_tma`.\n\n![Scatter Plot of image_width and image_height colored by is_tma](https://storage.googleapis.com/kaggle-forum-message-attachments/2481282/19785/2023-10-14%20110601.png)\n\n\nFor more details, please refer to the notebook below.\n\nhttps://www.kaggle.com/isakatsuyoshi/perfect-prediction-of-is-tma",
      "votes": 10
    },
    {
      "id": 2482468,
      "postDate": "2023-10-15T04:28:45.163Z",
      "content": "<p>I don't know if we can predict hidden test set TMAs like this. There could be some edge cases that are larger than WSIs. I also don't think we should predict whether it is WSI or TMA. We should try to build a model that could generalize to both image types.</p>",
      "rawMarkdown": "I don't know if we can predict hidden test set TMAs like this. There could be some edge cases that are larger than WSIs. I also don't think we should predict whether it is WSI or TMA. We should try to build a model that could generalize to both image types.",
      "votes": 6
    },
    {
      "id": 2482242,
      "postDate": "2023-10-14T19:32:01.870Z",
      "content": "<p>The goal is not to predict whether it is a TMA or not but instead to predict the type of cancer.  TMA is just a specific image type at 40x magnification.  This could be useful, however, as the authors stated that TMA is typically around 4,000x4,000</p>",
      "rawMarkdown": "The goal is not to predict whether it is a TMA or not but instead to predict the type of cancer.  TMA is just a specific image type at 40x magnification.  This could be useful, however, as the authors stated that TMA is typically around 4,000x4,000",
      "votes": 4,
      "replies": [
        {
          "id": 2482452,
          "postDate": "2023-10-15T04:05:49.740Z",
          "content": "<p><a href=\"https://www.kaggle.com/connor\" target=\"_blank\">@connor</a><br>\nThanks for the comment! Yes, you're right. I've done some exploratory data analysis just to double-check the data.</p>",
          "rawMarkdown": "@connor\nThanks for the comment! Yes, you're right. I've done some exploratory data analysis just to double-check the data.",
          "votes": 1
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 2482468,
      "author_name": "Gunes Evitan",
      "author_url": "",
      "post_date": "2023-10-15T04:28:45.163000",
      "content": "<p>I don't know if we can predict hidden test set TMAs like this. There could be some edge cases that are larger than WSIs. I also don't think we should predict whether it is WSI or TMA. We should try to build a model that could generalize to both image types.</p>",
      "votes": 6,
      "replies": []
    },
    {
      "id": 2482242,
      "author_name": "Connor",
      "author_url": "",
      "post_date": "2023-10-14T19:32:01.870000",
      "content": "<p>The goal is not to predict whether it is a TMA or not but instead to predict the type of cancer.  TMA is just a specific image type at 40x magnification.  This could be useful, however, as the authors stated that TMA is typically around 4,000x4,000</p>",
      "votes": 4,
      "replies": [
        {
          "id": 2482452,
          "author_name": "ISAKA Tsuyoshi",
          "author_url": "",
          "post_date": "2023-10-15T04:05:49.740000",
          "content": "<p><a href=\"https://www.kaggle.com/connor\" target=\"_blank\">@connor</a><br>\nThanks for the comment! Yes, you're right. I've done some exploratory data analysis just to double-check the data.</p>",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2481282": "Using a threshold value of around 5000 for both `image_width` and `image_height` seems to perfectly predict the value of `is_tma`.\n\n![Scatter Plot of image_width and image_height colored by is_tma](https://storage.googleapis.com/kaggle-forum-message-attachments/2481282/19785/2023-10-14%20110601.png)\n\n\nFor more details, please refer to the notebook below.\n\nhttps://www.kaggle.com/isakatsuyoshi/perfect-prediction-of-is-tma",
    "2482468": "I don't know if we can predict hidden test set TMAs like this. There could be some edge cases that are larger than WSIs. I also don't think we should predict whether it is WSI or TMA. We should try to build a model that could generalize to both image types.",
    "2482242": "The goal is not to predict whether it is a TMA or not but instead to predict the type of cancer.  TMA is just a specific image type at 40x magnification.  This could be useful, however, as the authors stated that TMA is typically around 4,000x4,000"
  }
}