{
  "id": 164514,
  "title": "How to cross 0.95",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/164514",
  "author_name": "",
  "post_date": "2020-07-06T15:32:42.024201300Z",
  "votes": 2,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Please give some ideas to cross 0.95  </p>",
  "messages": [
    {
      "id": "917527",
      "postDate": "07/06/2020 15:32:42",
      "content": "<p>Please give some ideas to cross 0.95  </p>",
      "rawMarkdown": "Please give some ideas to cross 0.95",
      "votes": null
    },
    {
      "id": "917542",
      "postDate": "07/06/2020 15:49:12",
      "content": "<p>Why aim for just 0.95, when you can target 0.960 😄 </p>\n\n<p><a href=\"/cdeotte\">@cdeotte</a>  already posted many helpful tips - see <a href=\"https://www.kaggle.com/c/siim-isic-melanoma-classification/discussion/164092\"><strong>Verified LB 0.960+</strong></a> </p>",
      "rawMarkdown": "Why aim for just 0.95, when you can target 0.960 😄 \n\n@cdeotte  already posted many helpful tips - see [**Verified LB 0.960+**](https://www.kaggle.com/c/siim-isic-melanoma-classification/discussion/164092)",
      "votes": null
    },
    {
      "id": "917545",
      "postDate": "07/06/2020 15:52:14",
      "content": "<p><a href=\"/sirishks\">@sirishks</a> what type of technics are you using? </p>",
      "rawMarkdown": "sirishks what type of technics are you using?",
      "votes": null
    },
    {
      "id": "917561",
      "postDate": "07/06/2020 15:56:09",
      "content": "<p>Hi. I don’t know what you are currently doing but here’s roughly what I’m doing.</p>\n\n<p>Use Metadata\nModel at least 3 different image sizes\nUse external data, the lb score will be lower \nEnsemble all the above.</p>\n\n<p>It’s the ensemble part that will make the difference if you can get the weights of the different models right.</p>\n\n<p>Have fun</p>",
      "rawMarkdown": "Hi. I don’t know what you are currently doing but here’s roughly what I’m doing.\n\nUse Metadata\nModel at least 3 different image sizes\nUse external data, the lb score will be lower \nEnsemble all the above.\n\nIt’s the ensemble part that will make the difference if you can get the weights of the different models right.\n\nHave fun",
      "votes": null
    },
    {
      "id": "917562",
      "postDate": "07/06/2020 15:56:35",
      "content": "<p>Just read all the Discussion posts!</p>\n\n<ol>\n<li>I am also trying all the ideas from public kernels and discussions</li>\n<li>Trying other new ideas from research papers (also taken from Discussion posts)</li>\n</ol>",
      "rawMarkdown": "Just read all the Discussion posts!\n\n1. I am also trying all the ideas from public kernels and discussions\n2. Trying other new ideas from research papers (also taken from Discussion posts)",
      "votes": null
    },
    {
      "id": "917568",
      "postDate": "07/06/2020 15:58:45",
      "content": "<p><a href=\"/mutantspore\">@mutantspore</a> thank you so much 😊</p>",
      "rawMarkdown": "mutantspore thank you so much 😊",
      "votes": null
    },
    {
      "id": "917569",
      "postDate": "07/06/2020 15:59:24",
      "content": "<p><a href=\"/sirishks\">@sirishks</a> Thank you so much, any tips and tricks, what type of models are you using? </p>",
      "rawMarkdown": "sirishks Thank you so much, any tips and tricks, what type of models are you using?",
      "votes": null
    },
    {
      "id": "917611",
      "postDate": "07/06/2020 16:27:28",
      "content": "<p>I am still <em>trying</em> a lot of things.</p>\n\n<p>Once I am able to structure my final ideas into a presentable format, then I will post the details (hopefully along with a Kernel), tentatively in the first week of August.</p>",
      "rawMarkdown": "I am still *trying* a lot of things.\n\nOnce I am able to structure my final ideas into a presentable format, then I will post the details (hopefully along with a Kernel), tentatively in the first week of August.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 917542,
      "author_name": "sirishks",
      "author_url": "",
      "post_date": "07/06/2020 15:49:12",
      "content": "<p>Why aim for just 0.95, when you can target 0.960 😄 </p>\n\n<p><a href=\"/cdeotte\">@cdeotte</a>  already posted many helpful tips - see <a href=\"https://www.kaggle.com/c/siim-isic-melanoma-classification/discussion/164092\"><strong>Verified LB 0.960+</strong></a> </p>",
      "votes": null,
      "replies": [
        {
          "id": 917545,
          "author_name": "gopidurgaprasad",
          "author_url": "",
          "post_date": "07/06/2020 15:52:14",
          "content": "<p><a href=\"/sirishks\">@sirishks</a> what type of technics are you using? </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 917562,
          "author_name": "sirishks",
          "author_url": "",
          "post_date": "07/06/2020 15:56:35",
          "content": "<p>Just read all the Discussion posts!</p>\n\n<ol>\n<li>I am also trying all the ideas from public kernels and discussions</li>\n<li>Trying other new ideas from research papers (also taken from Discussion posts)</li>\n</ol>",
          "votes": null,
          "replies": []
        },
        {
          "id": 917569,
          "author_name": "gopidurgaprasad",
          "author_url": "",
          "post_date": "07/06/2020 15:59:24",
          "content": "<p><a href=\"/sirishks\">@sirishks</a> Thank you so much, any tips and tricks, what type of models are you using? </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 917611,
          "author_name": "sirishks",
          "author_url": "",
          "post_date": "07/06/2020 16:27:28",
          "content": "<p>I am still <em>trying</em> a lot of things.</p>\n\n<p>Once I am able to structure my final ideas into a presentable format, then I will post the details (hopefully along with a Kernel), tentatively in the first week of August.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 917561,
      "author_name": "mutantspore",
      "author_url": "",
      "post_date": "07/06/2020 15:56:09",
      "content": "<p>Hi. I don’t know what you are currently doing but here’s roughly what I’m doing.</p>\n\n<p>Use Metadata\nModel at least 3 different image sizes\nUse external data, the lb score will be lower \nEnsemble all the above.</p>\n\n<p>It’s the ensemble part that will make the difference if you can get the weights of the different models right.</p>\n\n<p>Have fun</p>",
      "votes": null,
      "replies": [
        {
          "id": 917568,
          "author_name": "gopidurgaprasad",
          "author_url": "",
          "post_date": "07/06/2020 15:58:45",
          "content": "<p><a href=\"/mutantspore\">@mutantspore</a> thank you so much 😊</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "917527": "Please give some ideas to cross 0.95",
    "917542": "Why aim for just 0.95, when you can target 0.960 😄 \n\n@cdeotte  already posted many helpful tips - see [**Verified LB 0.960+**](https://www.kaggle.com/c/siim-isic-melanoma-classification/discussion/164092)",
    "917545": "sirishks what type of technics are you using?",
    "917561": "Hi. I don’t know what you are currently doing but here’s roughly what I’m doing.\n\nUse Metadata\nModel at least 3 different image sizes\nUse external data, the lb score will be lower \nEnsemble all the above.\n\nIt’s the ensemble part that will make the difference if you can get the weights of the different models right.\n\nHave fun",
    "917562": "Just read all the Discussion posts!\n\n1. I am also trying all the ideas from public kernels and discussions\n2. Trying other new ideas from research papers (also taken from Discussion posts)",
    "917568": "mutantspore thank you so much 😊",
    "917569": "sirishks Thank you so much, any tips and tricks, what type of models are you using?",
    "917611": "I am still *trying* a lot of things.\n\nOnce I am able to structure my final ideas into a presentable format, then I will post the details (hopefully along with a Kernel), tentatively in the first week of August."
  },
  "source": "meta"
}