{
  "id": 68371,
  "title": "Question about Targets",
  "url": "/competitions/human-protein-atlas-image-classification/discussion/68371",
  "author_name": "Shapes",
  "post_date": "2018-10-12T01:57:10.091000",
  "votes": 3,
  "comment_count": 4,
  "views": 0,
  "content": "<p>So the targets in general have a format like [24 7 2 0], starting with the highest number and ending with the lowest.\nBut I saw at least a couple instances where the lowest numbers come first: [0 8 10] for example.</p>\n\n<p>I wanted to check if this order is important for the evaluation of the submissions? I imagine most implementations will have predictions be translated from a \"many hot\" [0 1 0 .... 1 0] (28 values) format to the final one, and I'm wondering if this is going to affect the results.</p>\n\n<p>Or, for the computation of F1-macro, do the files get converted back into \"many-hot\"?</p>\n\n<p>Thanks in advance and sorry if the question is stupid.</p>",
  "messages": [
    {
      "id": 402820,
      "postDate": "2018-10-12T11:51:54.973Z",
      "content": "<p>Hello!</p>\n\n<p>The scoring is implemented in such a way that the order of 'Target' should not have an impact.</p>",
      "rawMarkdown": "Hello!\n\nThe scoring is implemented in such a way that the order of 'Target' should not have an impact.",
      "votes": 3,
      "replies": [
        {
          "id": 403655,
          "postDate": "2018-10-14T08:48:25.127Z",
          "content": "<p>Thank you for the prompt answer!</p>",
          "rawMarkdown": "Thank you for the prompt answer!"
        }
      ]
    },
    {
      "id": 402611,
      "postDate": "2018-10-12T01:57:10.090Z",
      "content": "<p>So the targets in general have a format like [24 7 2 0], starting with the highest number and ending with the lowest.\nBut I saw at least a couple instances where the lowest numbers come first: [0 8 10] for example.</p>\n\n<p>I wanted to check if this order is important for the evaluation of the submissions? I imagine most implementations will have predictions be translated from a \"many hot\" [0 1 0 .... 1 0] (28 values) format to the final one, and I'm wondering if this is going to affect the results.</p>\n\n<p>Or, for the computation of F1-macro, do the files get converted back into \"many-hot\"?</p>\n\n<p>Thanks in advance and sorry if the question is stupid.</p>",
      "rawMarkdown": "So the targets in general have a format like [24 7 2 0], starting with the highest number and ending with the lowest.\nBut I saw at least a couple instances where the lowest numbers come first: [0 8 10] for example.\n\nI wanted to check if this order is important for the evaluation of the submissions? I imagine most implementations will have predictions be translated from a \"many hot\" [0 1 0 .... 1 0] (28 values) format to the final one, and I'm wondering if this is going to affect the results.\n\nOr, for the computation of F1-macro, do the files get converted back into \"many-hot\"?\n\nThanks in advance and sorry if the question is stupid.\n\n",
      "votes": 3
    },
    {
      "id": 406945,
      "postDate": "2018-10-20T04:13:42.920Z",
      "content": "<p>I had the same question. It's maybe determined by the order of a result of sorting the table by class names.</p>\n\n<p>12   Actin filaments <br>\n24   Aggresome <br>\n22   Cell junctions <br>\n19   Centrosome <br>\n16   Cytokinetic bridge\n26   Cytoplasmic bodies\n25   Cytosol <br>\n13   Focal adhesion sites <br>\n11   Intermediate filaments\n20   Lipid droplets <br>\n10   Lysosomes <br>\n15   Microtubule ends <br>\n18   Microtubule organizing center\n14   Microtubules <br>\n23   Mitochondria <br>\n17   Mitotic spindle <br>\n21   Plasma membrane <br>\n27   Rods &amp; rings <br>\n6   Endoplasmic reticulum\n9   Endosomes <br>\n7   Golgi apparatus <br>\n5   Nuclear bodies <br>\n1   Nuclear membrane <br>\n4   Nuclear speckles <br>\n2   Nucleoli <br>\n3   Nucleoli fibrillar center <br>\n0   Nucleoplasm <br>\n8   Peroxisomes   </p>",
      "rawMarkdown": "I had the same question. It's maybe determined by the order of a result of sorting the table by class names.\n\n12\t Actin filaments   \n24\t Aggresome   \n22\t Cell junctions   \n19\t Centrosome   \n16\t Cytokinetic bridge\n26\t Cytoplasmic bodies\n25\t Cytosol   \n13\t Focal adhesion sites  \n11\t Intermediate filaments\n20\t Lipid droplets   \n10\t Lysosomes   \n15\t Microtubule ends  \n18\t Microtubule organizing center\n14\t Microtubules   \n23\t Mitochondria   \n17\t Mitotic spindle   \n21\t Plasma membrane   \n27\t Rods &amp; rings  \n6\tEndoplasmic reticulum\n9\tEndosomes   \n7\tGolgi apparatus   \n5\tNuclear bodies   \n1\tNuclear membrane   \n4\tNuclear speckles   \n2\tNucleoli   \n3\tNucleoli fibrillar center    \n0\tNucleoplasm  \n8\tPeroxisomes   ",
      "votes": 1
    },
    {
      "id": 403653,
      "postDate": "2018-10-14T08:47:50.843Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 402820,
      "author_name": "Casper Winsnes",
      "author_url": "",
      "post_date": "2018-10-12T11:51:54.973000",
      "content": "<p>Hello!</p>\n\n<p>The scoring is implemented in such a way that the order of 'Target' should not have an impact.</p>",
      "votes": 3,
      "replies": [
        {
          "id": 403655,
          "author_name": "Shapes",
          "author_url": "",
          "post_date": "2018-10-14T08:48:25.127000",
          "content": "<p>Thank you for the prompt answer!</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 406945,
      "author_name": "tamref",
      "author_url": "",
      "post_date": "2018-10-20T04:13:42.920000",
      "content": "<p>I had the same question. It's maybe determined by the order of a result of sorting the table by class names.</p>\n\n<p>12   Actin filaments <br>\n24   Aggresome <br>\n22   Cell junctions <br>\n19   Centrosome <br>\n16   Cytokinetic bridge\n26   Cytoplasmic bodies\n25   Cytosol <br>\n13   Focal adhesion sites <br>\n11   Intermediate filaments\n20   Lipid droplets <br>\n10   Lysosomes <br>\n15   Microtubule ends <br>\n18   Microtubule organizing center\n14   Microtubules <br>\n23   Mitochondria <br>\n17   Mitotic spindle <br>\n21   Plasma membrane <br>\n27   Rods &amp; rings <br>\n6   Endoplasmic reticulum\n9   Endosomes <br>\n7   Golgi apparatus <br>\n5   Nuclear bodies <br>\n1   Nuclear membrane <br>\n4   Nuclear speckles <br>\n2   Nucleoli <br>\n3   Nucleoli fibrillar center <br>\n0   Nucleoplasm <br>\n8   Peroxisomes   </p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 403653,
      "author_name": "",
      "author_url": "",
      "post_date": "2018-10-14T08:47:50.843000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "402820": "Hello!\n\nThe scoring is implemented in such a way that the order of 'Target' should not have an impact.",
    "402611": "So the targets in general have a format like [24 7 2 0], starting with the highest number and ending with the lowest.\nBut I saw at least a couple instances where the lowest numbers come first: [0 8 10] for example.\n\nI wanted to check if this order is important for the evaluation of the submissions? I imagine most implementations will have predictions be translated from a \"many hot\" [0 1 0 .... 1 0] (28 values) format to the final one, and I'm wondering if this is going to affect the results.\n\nOr, for the computation of F1-macro, do the files get converted back into \"many-hot\"?\n\nThanks in advance and sorry if the question is stupid.\n\n",
    "406945": "I had the same question. It's maybe determined by the order of a result of sorting the table by class names.\n\n12\t Actin filaments   \n24\t Aggresome   \n22\t Cell junctions   \n19\t Centrosome   \n16\t Cytokinetic bridge\n26\t Cytoplasmic bodies\n25\t Cytosol   \n13\t Focal adhesion sites  \n11\t Intermediate filaments\n20\t Lipid droplets   \n10\t Lysosomes   \n15\t Microtubule ends  \n18\t Microtubule organizing center\n14\t Microtubules   \n23\t Mitochondria   \n17\t Mitotic spindle   \n21\t Plasma membrane   \n27\t Rods &amp; rings  \n6\tEndoplasmic reticulum\n9\tEndosomes   \n7\tGolgi apparatus   \n5\tNuclear bodies   \n1\tNuclear membrane   \n4\tNuclear speckles   \n2\tNucleoli   \n3\tNucleoli fibrillar center    \n0\tNucleoplasm  \n8\tPeroxisomes   ",
    "403653": ""
  }
}