{
  "id": 135068,
  "title": "Target label 32K. Whats the best Idea?",
  "url": "/competitions/herbarium-2020-fgvc7/discussion/135068",
  "author_name": "",
  "post_date": "2020-03-11T22:08:02.310194500Z",
  "votes": 2,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I have seen that data has 32k unique target label. I cannot remember how to approach it exactly, can any one have idea.</p>",
  "messages": [
    {
      "id": "769419",
      "postDate": "03/11/2020 22:08:02",
      "content": "<p>I have seen that data has 32k unique target label. I cannot remember how to approach it exactly, can any one have idea.</p>",
      "rawMarkdown": "I have seen that data has 32k unique target label. I cannot remember how to approach it exactly, can any one have idea.",
      "votes": null
    },
    {
      "id": "770125",
      "postDate": "03/12/2020 15:43:26",
      "content": "<p>For my mind, the best idea in this situation is used Siamese Neural Net with triplet loss.  </p>",
      "rawMarkdown": "For my mind, the best idea in this situation is used Siamese Neural Net with triplet loss.",
      "votes": null
    },
    {
      "id": "776118",
      "postDate": "03/17/2020 06:08:47",
      "content": "<p>Hierarchical Softmax?\n<a href=\"https://leimao.github.io/article/Hierarchical-Softmax/\">https://leimao.github.io/article/Hierarchical-Softmax/</a></p>",
      "rawMarkdown": "Hierarchical Softmax?\nhttps://leimao.github.io/article/Hierarchical-Softmax/",
      "votes": null
    },
    {
      "id": "790882",
      "postDate": "03/29/2020 23:18:42",
      "content": "<p><a href=\"/khursani8\">@khursani8</a> thanks for the share.</p>",
      "rawMarkdown": "khursani8 thanks for the share.",
      "votes": null
    },
    {
      "id": "790883",
      "postDate": "03/29/2020 23:19:16",
      "content": "<p>Great i will look into it, Thanks <a href=\"/miklgr500\">@miklgr500</a> </p>",
      "rawMarkdown": "Great i will look into it, Thanks @miklgr500",
      "votes": null
    },
    {
      "id": "806386",
      "postDate": "04/13/2020 17:17:24",
      "content": "<p>Does it work? Have you tried it? Have you tried it on other data sets? </p>",
      "rawMarkdown": "Does it work? Have you tried it? Have you tried it on other data sets?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 770125,
      "author_name": "miklgr500",
      "author_url": "",
      "post_date": "03/12/2020 15:43:26",
      "content": "<p>For my mind, the best idea in this situation is used Siamese Neural Net with triplet loss.  </p>",
      "votes": null,
      "replies": [
        {
          "id": 790883,
          "author_name": "kalyankkr",
          "author_url": "",
          "post_date": "03/29/2020 23:19:16",
          "content": "<p>Great i will look into it, Thanks <a href=\"/miklgr500\">@miklgr500</a> </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 776118,
      "author_name": "khursani8",
      "author_url": "",
      "post_date": "03/17/2020 06:08:47",
      "content": "<p>Hierarchical Softmax?\n<a href=\"https://leimao.github.io/article/Hierarchical-Softmax/\">https://leimao.github.io/article/Hierarchical-Softmax/</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 790882,
          "author_name": "kalyankkr",
          "author_url": "",
          "post_date": "03/29/2020 23:18:42",
          "content": "<p><a href=\"/khursani8\">@khursani8</a> thanks for the share.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 806386,
          "author_name": "valanm",
          "author_url": "",
          "post_date": "04/13/2020 17:17:24",
          "content": "<p>Does it work? Have you tried it? Have you tried it on other data sets? </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "769419": "I have seen that data has 32k unique target label. I cannot remember how to approach it exactly, can any one have idea.",
    "770125": "For my mind, the best idea in this situation is used Siamese Neural Net with triplet loss.",
    "776118": "Hierarchical Softmax?\nhttps://leimao.github.io/article/Hierarchical-Softmax/",
    "790882": "khursani8 thanks for the share.",
    "790883": "Great i will look into it, Thanks @miklgr500",
    "806386": "Does it work? Have you tried it? Have you tried it on other data sets?"
  },
  "source": "meta"
}