{
  "id": 110280,
  "title": "\"ANY\" manipulation",
  "url": "/competitions/rsna-intracranial-hemorrhage-detection/discussion/110280",
  "author_name": "Gurgel",
  "post_date": "2019-09-26T14:33:15.297000",
  "votes": 0,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I noticed my model isn't giving \"any\" the value it should have, for example if the biggest label value for a certain ID is epidural (= 0.7, for example), shouldn't any also have this value? </p>\n\n<p>Should we manually adjust the \"any\" value to be equal to the highest one from the other 5 classes? Has anyone tried it to improve their scores?</p>\n\n<p>Additionally, are u training your model with 5 or 6 labels (excluding any from training and manually putting it's value after)?</p>",
  "messages": [
    {
      "id": 634645,
      "postDate": "2019-09-26T14:58:01.010Z",
      "content": "<p>Why do you expect 'any' to be the the max of the other values?</p>\n\n<p>If the value for class1 is 0.7 and class2 is 0.4, 'any' should be higher then when class1 is 0.7 and class2 is 0. </p>",
      "rawMarkdown": "Why do you expect 'any' to be the the max of the other values?\n\nIf the value for class1 is 0.7 and class2 is 0.4, 'any' should be higher then when class1 is 0.7 and class2 is 0. ",
      "votes": 3,
      "replies": [
        {
          "id": 634658,
          "postDate": "2019-09-26T15:09:03.623Z",
          "content": "<p>You are right. Rephrasing it: 'any' should be AT LEAST the maximum of these values. But I think I made a mistake when looking at the results and I thought the 'any' value was not being greater than the largest one. Will take a better look to see if it always follows this rule.</p>",
          "rawMarkdown": "You are right. Rephrasing it: 'any' should be AT LEAST the maximum of these values. But I think I made a mistake when looking at the results and I thought the 'any' value was not being greater than the largest one. Will take a better look to see if it always follows this rule."
        }
      ]
    },
    {
      "id": 634629,
      "postDate": "2019-09-26T14:33:15.297Z",
      "content": "<p>I noticed my model isn't giving \"any\" the value it should have, for example if the biggest label value for a certain ID is epidural (= 0.7, for example), shouldn't any also have this value? </p>\n\n<p>Should we manually adjust the \"any\" value to be equal to the highest one from the other 5 classes? Has anyone tried it to improve their scores?</p>\n\n<p>Additionally, are u training your model with 5 or 6 labels (excluding any from training and manually putting it's value after)?</p>",
      "rawMarkdown": "I noticed my model isn't giving \"any\" the value it should have, for example if the biggest label value for a certain ID is epidural (= 0.7, for example), shouldn't any also have this value? \n\nShould we manually adjust the \"any\" value to be equal to the highest one from the other 5 classes? Has anyone tried it to improve their scores?\n\nAdditionally, are u training your model with 5 or 6 labels (excluding any from training and manually putting it's value after)?\n"
    }
  ],
  "comments": [
    {
      "id": 634645,
      "author_name": "yuval reina",
      "author_url": "",
      "post_date": "2019-09-26T14:58:01.010000",
      "content": "<p>Why do you expect 'any' to be the the max of the other values?</p>\n\n<p>If the value for class1 is 0.7 and class2 is 0.4, 'any' should be higher then when class1 is 0.7 and class2 is 0. </p>",
      "votes": 3,
      "replies": [
        {
          "id": 634658,
          "author_name": "Gurgel",
          "author_url": "",
          "post_date": "2019-09-26T15:09:03.623000",
          "content": "<p>You are right. Rephrasing it: 'any' should be AT LEAST the maximum of these values. But I think I made a mistake when looking at the results and I thought the 'any' value was not being greater than the largest one. Will take a better look to see if it always follows this rule.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "634645": "Why do you expect 'any' to be the the max of the other values?\n\nIf the value for class1 is 0.7 and class2 is 0.4, 'any' should be higher then when class1 is 0.7 and class2 is 0. ",
    "634629": "I noticed my model isn't giving \"any\" the value it should have, for example if the biggest label value for a certain ID is epidural (= 0.7, for example), shouldn't any also have this value? \n\nShould we manually adjust the \"any\" value to be equal to the highest one from the other 5 classes? Has anyone tried it to improve their scores?\n\nAdditionally, are u training your model with 5 or 6 labels (excluding any from training and manually putting it's value after)?\n"
  }
}