{
  "id": 211194,
  "title": "one dark magic trick",
  "url": "/competitions/ranzcr-clip-catheter-line-classification/discussion/211194",
  "author_name": "",
  "post_date": "2021-01-14T06:23:06.794094900Z",
  "votes": 51,
  "comment_count": 6,
  "views": 0,
  "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F7352618c5477245f0459a435b8011a1c%2FSelection_038.png?generation=1610605307205641&amp;alt=media\" alt=\"\"></p>\n<p>because of the distribution of the results shown above, try this at your validation and public test set:</p>\n<pre><code>ensemble += p[i]\nensemble = ensemble/num_of_model\n\nchange to\n\nensemble += p[i]**0.5\nensemble = ensemble/num_of_model\n</code></pre>\n<p>aka. hand-adjusting biasing …</p>",
  "messages": [
    {
      "id": "1152429",
      "postDate": "01/14/2021 06:23:06",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F7352618c5477245f0459a435b8011a1c%2FSelection_038.png?generation=1610605307205641&amp;alt=media\" alt=\"\"></p>\n<p>because of the distribution of the results shown above, try this at your validation and public test set:</p>\n<pre><code>ensemble += p[i]\nensemble = ensemble/num_of_model\n\nchange to\n\nensemble += p[i]**0.5\nensemble = ensemble/num_of_model\n</code></pre>\n<p>aka. hand-adjusting biasing …</p>",
      "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F7352618c5477245f0459a435b8011a1c%2FSelection_038.png?generation=1610605307205641&alt=media)\n\nbecause of the distribution of the results shown above, try this at your validation and public test set:\n\n```\nensemble += p[i]\nensemble = ensemble/num_of_model\n\nchange to\n\nensemble += p[i]**0.5\nensemble = ensemble/num_of_model\n\n```\n\naka. hand-adjusting biasing ...",
      "votes": null
    },
    {
      "id": "1152430",
      "postDate": "01/14/2021 06:26:08",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F6e579a26dfd67f4b8f413248c953fb40%2FSelection_039.png?generation=1610605505200399&amp;alt=media\" alt=\"\"></p>\n<p>x-axis (prediction from one model, e.g. resnet200d)<br>\ny-axis (prediction from another model, e.g. efficienetb5)</p>\n<p>blue: pos samples<br>\norange: neg samples</p>",
      "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F6e579a26dfd67f4b8f413248c953fb40%2FSelection_039.png?generation=1610605505200399&alt=media)\n\nx-axis (prediction from one model, e.g. resnet200d)\ny-axis (prediction from another model, e.g. efficienetb5)\n\nblue: pos samples\norange: neg samples",
      "votes": null
    },
    {
      "id": "1152584",
      "postDate": "01/14/2021 09:37:57",
      "content": "<p>Thanks - it indeed provides small uplift on CV</p>",
      "rawMarkdown": "Thanks - it indeed provides small uplift on CV",
      "votes": null
    },
    {
      "id": "1154056",
      "postDate": "01/15/2021 11:23:45",
      "content": "<p>Thanks a lot <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> are you talking about 145 validation cases?</p>\n<p>Could you explain please the meaning of orange, green and blue cases?</p>\n<p>As far as I understood the green ones aren't tagged as CVC. and I'm not very sure about the meaning of pos/neg in this context.</p>\n<p>I must be missing something</p>",
      "rawMarkdown": "Thanks a lot @hengck23 are you talking about 145 validation cases?\n\nCould you explain please the meaning of orange, green and blue cases?\n\nAs far as I understood the green ones aren't tagged as CVC. and I'm not very sure about the meaning of pos/neg in this context.\n\nI must be missing something",
      "votes": null
    },
    {
      "id": "1155035",
      "postDate": "01/16/2021 06:31:42",
      "content": "<p>here is an example:<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Fafe732b0a5689f7522b3044867e855d5%2FSelection_050.png?generation=1610778700753031&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "here is an example:\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Fafe732b0a5689f7522b3044867e855d5%2FSelection_050.png?generation=1610778700753031&alt=media)",
      "votes": null
    },
    {
      "id": "1155193",
      "postDate": "01/16/2021 09:21:52",
      "content": "<p>thanks a lot <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> !</p>",
      "rawMarkdown": "thanks a lot @hengck23 !",
      "votes": null
    },
    {
      "id": "1162379",
      "postDate": "01/21/2021 05:47:38",
      "content": "<p><a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> please confirm and help understand your approach; are you saying to take weighted average of two model probability prediction and prepare submission file?</p>",
      "rawMarkdown": "hengck23 please confirm and help understand your approach; are you saying to take weighted average of two model probability prediction and prepare submission file?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1152430,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "01/14/2021 06:26:08",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F6e579a26dfd67f4b8f413248c953fb40%2FSelection_039.png?generation=1610605505200399&amp;alt=media\" alt=\"\"></p>\n<p>x-axis (prediction from one model, e.g. resnet200d)<br>\ny-axis (prediction from another model, e.g. efficienetb5)</p>\n<p>blue: pos samples<br>\norange: neg samples</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1152584,
      "author_name": "raddar",
      "author_url": "",
      "post_date": "01/14/2021 09:37:57",
      "content": "<p>Thanks - it indeed provides small uplift on CV</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1154056,
      "author_name": "virilo",
      "author_url": "",
      "post_date": "01/15/2021 11:23:45",
      "content": "<p>Thanks a lot <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> are you talking about 145 validation cases?</p>\n<p>Could you explain please the meaning of orange, green and blue cases?</p>\n<p>As far as I understood the green ones aren't tagged as CVC. and I'm not very sure about the meaning of pos/neg in this context.</p>\n<p>I must be missing something</p>",
      "votes": null,
      "replies": [
        {
          "id": 1155035,
          "author_name": "hengck23",
          "author_url": "",
          "post_date": "01/16/2021 06:31:42",
          "content": "<p>here is an example:<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Fafe732b0a5689f7522b3044867e855d5%2FSelection_050.png?generation=1610778700753031&amp;alt=media\" alt=\"\"></p>",
          "votes": null,
          "replies": [
            {
              "id": 1155193,
              "author_name": "virilo",
              "author_url": "",
              "post_date": "01/16/2021 09:21:52",
              "content": "<p>thanks a lot <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> !</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 1162379,
      "author_name": "chandraroy",
      "author_url": "",
      "post_date": "01/21/2021 05:47:38",
      "content": "<p><a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> please confirm and help understand your approach; are you saying to take weighted average of two model probability prediction and prepare submission file?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1152429": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F7352618c5477245f0459a435b8011a1c%2FSelection_038.png?generation=1610605307205641&alt=media)\n\nbecause of the distribution of the results shown above, try this at your validation and public test set:\n\n```\nensemble += p[i]\nensemble = ensemble/num_of_model\n\nchange to\n\nensemble += p[i]**0.5\nensemble = ensemble/num_of_model\n\n```\n\naka. hand-adjusting biasing ...",
    "1152430": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F6e579a26dfd67f4b8f413248c953fb40%2FSelection_039.png?generation=1610605505200399&alt=media)\n\nx-axis (prediction from one model, e.g. resnet200d)\ny-axis (prediction from another model, e.g. efficienetb5)\n\nblue: pos samples\norange: neg samples",
    "1152584": "Thanks - it indeed provides small uplift on CV",
    "1154056": "Thanks a lot @hengck23 are you talking about 145 validation cases?\n\nCould you explain please the meaning of orange, green and blue cases?\n\nAs far as I understood the green ones aren't tagged as CVC. and I'm not very sure about the meaning of pos/neg in this context.\n\nI must be missing something",
    "1155035": "here is an example:\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2Fafe732b0a5689f7522b3044867e855d5%2FSelection_050.png?generation=1610778700753031&alt=media)",
    "1155193": "thanks a lot @hengck23 !",
    "1162379": "hengck23 please confirm and help understand your approach; are you saying to take weighted average of two model probability prediction and prepare submission file?"
  },
  "source": "meta"
}