{
  "id": 109447,
  "title": "weighted multi-label logarithmic loss",
  "url": "/competitions/rsna-intracranial-hemorrhage-detection/discussion/109447",
  "author_name": "arnowaczynski",
  "post_date": "2019-09-19T08:08:30.131000",
  "votes": 24,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Why were the weight values in the evaluation metric not given to us?</p>",
  "messages": [
    {
      "id": 629822,
      "postDate": "2019-09-19T08:08:30.130Z",
      "content": "<p>Why were the weight values in the evaluation metric not given to us?</p>",
      "rawMarkdown": "Why were the weight values in the evaluation metric not given to us?",
      "votes": 24
    },
    {
      "id": 631881,
      "postDate": "2019-09-22T21:45:56.947Z",
      "content": "<p>I found the 'any' weight is x2 of others. <br>\n<a href=\"https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/109526#latest-630190\">https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/109526#latest-630190</a></p>\n\n<p>But any == max(epidural, intraparenchymal, intraventricular, subarachnoid, subdural), so I think training model without any label and fill any as post-processing is good.  </p>",
      "rawMarkdown": "I found the 'any' weight is x2 of others.  \nhttps://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/109526#latest-630190\n\nBut any == max(epidural, intraparenchymal, intraventricular, subarachnoid, subdural), so I think training model without any label and fill any as post-processing is good.  ",
      "votes": 3,
      "replies": [
        {
          "id": 631887,
          "postDate": "2019-09-22T21:54:54.030Z",
          "content": "<p>Awesome! Thanks for sharing!</p>",
          "rawMarkdown": "Awesome! Thanks for sharing!"
        }
      ]
    },
    {
      "id": 631866,
      "postDate": "2019-09-22T20:46:44.297Z",
      "content": "<p>It's very odd for a Kaggle competition to be so vague about the metric. I hope we get more information soon. For now maybe we can give a bit more weight to 'any' and count the rest as equal weight?</p>\n\n<p>For example:\n<code>\nweights = {epidural : 0.15, intraparenchymal : 0.15, \nintraventricular : 0.15, subarachnoid : 0.15, \nsubdural : 0.15, any : 0.25}\n</code></p>\n\n<p>What do you think?</p>",
      "rawMarkdown": "It's very odd for a Kaggle competition to be so vague about the metric. I hope we get more information soon. For now maybe we can give a bit more weight to 'any' and count the rest as equal weight?\n\nFor example:\n```\nweights = {epidural : 0.15, intraparenchymal : 0.15, \nintraventricular : 0.15, subarachnoid : 0.15, \nsubdural : 0.15, any : 0.25}\n```\n\nWhat do you think?\n\n",
      "votes": 1,
      "replies": [
        {
          "id": 631868,
          "postDate": "2019-09-22T20:47:30.190Z",
          "content": "<p>Or maybe even more weight to 'any'?</p>\n\n<p><code>\nweights = {epidural : 0.1, intraparenchymal : 0.1, \nintraventricular : 0.1, subarachnoid : 0.1, \nsubdural : 0.1, any : 0.4}\n</code></p>",
          "rawMarkdown": "Or maybe even more weight to 'any'?\n\n```\nweights = {epidural : 0.1, intraparenchymal : 0.1, \nintraventricular : 0.1, subarachnoid : 0.1, \nsubdural : 0.1, any : 0.4}\n```",
          "votes": 1
        }
      ]
    },
    {
      "id": 632745,
      "postDate": "2019-09-24T02:25:42.320Z",
      "rawMarkdown": "",
      "votes": -1,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 631881,
      "author_name": "kambarakun",
      "author_url": "",
      "post_date": "2019-09-22T21:45:56.947000",
      "content": "<p>I found the 'any' weight is x2 of others. <br>\n<a href=\"https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/109526#latest-630190\">https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/109526#latest-630190</a></p>\n\n<p>But any == max(epidural, intraparenchymal, intraventricular, subarachnoid, subdural), so I think training model without any label and fill any as post-processing is good.  </p>",
      "votes": 3,
      "replies": [
        {
          "id": 631887,
          "author_name": "Carlo Lepelaars",
          "author_url": "",
          "post_date": "2019-09-22T21:54:54.030000",
          "content": "<p>Awesome! Thanks for sharing!</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 631866,
      "author_name": "Carlo Lepelaars",
      "author_url": "",
      "post_date": "2019-09-22T20:46:44.297000",
      "content": "<p>It's very odd for a Kaggle competition to be so vague about the metric. I hope we get more information soon. For now maybe we can give a bit more weight to 'any' and count the rest as equal weight?</p>\n\n<p>For example:\n<code>\nweights = {epidural : 0.15, intraparenchymal : 0.15, \nintraventricular : 0.15, subarachnoid : 0.15, \nsubdural : 0.15, any : 0.25}\n</code></p>\n\n<p>What do you think?</p>",
      "votes": 1,
      "replies": [
        {
          "id": 631868,
          "author_name": "Carlo Lepelaars",
          "author_url": "",
          "post_date": "2019-09-22T20:47:30.190000",
          "content": "<p>Or maybe even more weight to 'any'?</p>\n\n<p><code>\nweights = {epidural : 0.1, intraparenchymal : 0.1, \nintraventricular : 0.1, subarachnoid : 0.1, \nsubdural : 0.1, any : 0.4}\n</code></p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 632745,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-09-24T02:25:42.320000",
      "content": "",
      "votes": -1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "629822": "Why were the weight values in the evaluation metric not given to us?",
    "631881": "I found the 'any' weight is x2 of others.  \nhttps://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/109526#latest-630190\n\nBut any == max(epidural, intraparenchymal, intraventricular, subarachnoid, subdural), so I think training model without any label and fill any as post-processing is good.  ",
    "631866": "It's very odd for a Kaggle competition to be so vague about the metric. I hope we get more information soon. For now maybe we can give a bit more weight to 'any' and count the rest as equal weight?\n\nFor example:\n```\nweights = {epidural : 0.15, intraparenchymal : 0.15, \nintraventricular : 0.15, subarachnoid : 0.15, \nsubdural : 0.15, any : 0.25}\n```\n\nWhat do you think?\n\n",
    "632745": ""
  }
}