{
  "id": 310242,
  "title": "Explanation of Evaluation index 'MAP@5'",
  "url": "/competitions/happy-whale-and-dolphin/discussion/310242",
  "author_name": "",
  "post_date": "2022-02-28T10:34:30.403343900Z",
  "votes": 14,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I posted the image about explanation of evaluation index in this competition, 'MAP@5'.<br>\nPlease refer to the rink below.<br>\n![<a href=\"https://user-images.githubusercontent.com/52144537/155967641-d9b89766-b171-4ede-bbc1-63ff9dbf24fd.png\" target=\"_blank\">https://user-images.githubusercontent.com/52144537/155967641-d9b89766-b171-4ede-bbc1-63ff9dbf24fd.png</a>] (Image link)</p>\n<p>please let me know if there are any mistakes.<br>\nthanks</p>",
  "messages": [
    {
      "id": "1707327",
      "postDate": "02/28/2022 10:34:30",
      "content": "<p>I posted the image about explanation of evaluation index in this competition, 'MAP@5'.<br>\nPlease refer to the rink below.<br>\n![<a href=\"https://user-images.githubusercontent.com/52144537/155967641-d9b89766-b171-4ede-bbc1-63ff9dbf24fd.png\" target=\"_blank\">https://user-images.githubusercontent.com/52144537/155967641-d9b89766-b171-4ede-bbc1-63ff9dbf24fd.png</a>] (Image link)</p>\n<p>please let me know if there are any mistakes.<br>\nthanks</p>",
      "rawMarkdown": "I posted the image about explanation of evaluation index in this competition, 'MAP@5'.\nPlease refer to the rink below.\n![https://user-images.githubusercontent.com/52144537/155967641-d9b89766-b171-4ede-bbc1-63ff9dbf24fd.png] (Image link)\n\nplease let me know if there are any mistakes.\nthanks",
      "votes": null
    },
    {
      "id": "1713319",
      "postDate": "03/05/2022 22:26:39",
      "content": "<p>For your info, image URLs are still disabled within Kaggle, so you can use this free image hosting service: <a href=\"https://imgbb.com/\" target=\"_blank\">https://imgbb.com/</a></p>\n<p>Thanks for the explanation. 👌 </p>",
      "rawMarkdown": "For your info, image URLs are still disabled within Kaggle, so you can use this free image hosting service: https://imgbb.com/\n\nThanks for the explanation. 👌",
      "votes": null
    },
    {
      "id": "1713603",
      "postDate": "03/06/2022 07:27:43",
      "content": "<p>Thanks for comments.<br>\nSo, You can not watch the image i posted??</p>",
      "rawMarkdown": "Thanks for comments.\nSo, You can not watch the image i posted??",
      "votes": null
    },
    {
      "id": "1713909",
      "postDate": "03/06/2022 13:20:21",
      "content": "<p>No it isn't inlined but it can be seen if I click on it of course. </p>",
      "rawMarkdown": "No it isn't inlined but it can be seen if I click on it of course.",
      "votes": null
    },
    {
      "id": "1729844",
      "postDate": "03/20/2022 15:38:21",
      "content": "<p>So basically, you get full credit if your first guess is correct, otherwise 1/2 if your second guess is correct, otherwise 1/3 for the third guess, …</p>\n<p>There are still a few things I find odd about this terminology:</p>\n<ul>\n<li>precision (e.g. for a search engine) normally penalizes you the larger the fraction of false positives you return.  In contrast, AP only penalizes you for false positives earlier in the list than the correct ID.   In particular, if you don't get the correct ID at all, none of your false positives hurt your score.</li>\n<li>the description of this <em>average</em> precision <em>at</em> 5 seems a bit misleading: it is really the <em>sum</em> of precision for all values of k <em>up to and including</em> 5.</li>\n</ul>",
      "rawMarkdown": "So basically, you get full credit if your first guess is correct, otherwise 1/2 if your second guess is correct, otherwise 1/3 for the third guess, ...\n\nThere are still a few things I find odd about this terminology:\n* precision (e.g. for a search engine) normally penalizes you the larger the fraction of false positives you return.  In contrast, AP only penalizes you for false positives earlier in the list than the correct ID.   In particular, if you don't get the correct ID at all, none of your false positives hurt your score.\n* the description of this *average* precision *at* 5 seems a bit misleading: it is really the *sum* of precision for all values of k *up to and including* 5.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1713319,
      "author_name": "yassinealouini",
      "author_url": "",
      "post_date": "03/05/2022 22:26:39",
      "content": "<p>For your info, image URLs are still disabled within Kaggle, so you can use this free image hosting service: <a href=\"https://imgbb.com/\" target=\"_blank\">https://imgbb.com/</a></p>\n<p>Thanks for the explanation. 👌 </p>",
      "votes": null,
      "replies": [
        {
          "id": 1713603,
          "author_name": "haruki741",
          "author_url": "",
          "post_date": "03/06/2022 07:27:43",
          "content": "<p>Thanks for comments.<br>\nSo, You can not watch the image i posted??</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1713909,
          "author_name": "yassinealouini",
          "author_url": "",
          "post_date": "03/06/2022 13:20:21",
          "content": "<p>No it isn't inlined but it can be seen if I click on it of course. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1729844,
      "author_name": "talk2dfox",
      "author_url": "",
      "post_date": "03/20/2022 15:38:21",
      "content": "<p>So basically, you get full credit if your first guess is correct, otherwise 1/2 if your second guess is correct, otherwise 1/3 for the third guess, …</p>\n<p>There are still a few things I find odd about this terminology:</p>\n<ul>\n<li>precision (e.g. for a search engine) normally penalizes you the larger the fraction of false positives you return.  In contrast, AP only penalizes you for false positives earlier in the list than the correct ID.   In particular, if you don't get the correct ID at all, none of your false positives hurt your score.</li>\n<li>the description of this <em>average</em> precision <em>at</em> 5 seems a bit misleading: it is really the <em>sum</em> of precision for all values of k <em>up to and including</em> 5.</li>\n</ul>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1707327": "I posted the image about explanation of evaluation index in this competition, 'MAP@5'.\nPlease refer to the rink below.\n![https://user-images.githubusercontent.com/52144537/155967641-d9b89766-b171-4ede-bbc1-63ff9dbf24fd.png] (Image link)\n\nplease let me know if there are any mistakes.\nthanks",
    "1713319": "For your info, image URLs are still disabled within Kaggle, so you can use this free image hosting service: https://imgbb.com/\n\nThanks for the explanation. 👌",
    "1713603": "Thanks for comments.\nSo, You can not watch the image i posted??",
    "1713909": "No it isn't inlined but it can be seen if I click on it of course.",
    "1729844": "So basically, you get full credit if your first guess is correct, otherwise 1/2 if your second guess is correct, otherwise 1/3 for the third guess, ...\n\nThere are still a few things I find odd about this terminology:\n* precision (e.g. for a search engine) normally penalizes you the larger the fraction of false positives you return.  In contrast, AP only penalizes you for false positives earlier in the list than the correct ID.   In particular, if you don't get the correct ID at all, none of your false positives hurt your score.\n* the description of this *average* precision *at* 5 seems a bit misleading: it is really the *sum* of precision for all values of k *up to and including* 5."
  },
  "source": "meta"
}