{
  "id": 114395,
  "title": "Hints for late joiners?",
  "url": "/competitions/understanding_cloud_organization/discussion/114395",
  "author_name": "",
  "post_date": "2019-10-26T00:07:56.066936200Z",
  "votes": 15,
  "comment_count": 11,
  "views": 0,
  "content": "<p>Dear all,</p>\n\n<p>Could you have any suggestions on where to start reading into this competition? Lots of discussions are about postprocessing. Not sure if that's a good starting point.</p>",
  "messages": [
    {
      "id": "658386",
      "postDate": "10/26/2019 00:07:56",
      "content": "<p>Dear all,</p>\n\n<p>Could you have any suggestions on where to start reading into this competition? Lots of discussions are about postprocessing. Not sure if that's a good starting point.</p>",
      "rawMarkdown": "Dear all,\n\nCould you have any suggestions on where to start reading into this competition? Lots of discussions are about postprocessing. Not sure if that's a good starting point.",
      "votes": null
    },
    {
      "id": "658632",
      "postDate": "10/26/2019 09:32:39",
      "content": "<p><a href=\"https://www.kaggle.com/c/understanding_cloud_organization/discussion/110650#latest-639010\">https://www.kaggle.com/c/understanding_cloud_organization/discussion/110650#latest-639010</a></p>\n\n<p><a href=\"https://www.kaggle.com/c/understanding_cloud_organization/discussion/104573#latest-635362\">https://www.kaggle.com/c/understanding_cloud_organization/discussion/104573#latest-635362</a></p>\n\n<p><a href=\"https://www.kaggle.com/c/understanding_cloud_organization/discussion/110303#latest-643328\">https://www.kaggle.com/c/understanding_cloud_organization/discussion/110303#latest-643328</a></p>\n\n<p><a href=\"https://www.kaggle.com/c/understanding_cloud_organization/discussion/108306#latest-631556\">https://www.kaggle.com/c/understanding_cloud_organization/discussion/108306#latest-631556</a></p>\n\n<p>but why you think postprocessing is not a good starting point?</p>",
      "rawMarkdown": "https://www.kaggle.com/c/understanding_cloud_organization/discussion/110650#latest-639010\n\nhttps://www.kaggle.com/c/understanding_cloud_organization/discussion/104573#latest-635362\n\nhttps://www.kaggle.com/c/understanding_cloud_organization/discussion/110303#latest-643328\n\nhttps://www.kaggle.com/c/understanding_cloud_organization/discussion/108306#latest-631556\n\nbut why you think postprocessing is not a good starting point?",
      "votes": null
    },
    {
      "id": "658905",
      "postDate": "10/26/2019 18:08:28",
      "content": "<p>Thanks. I am worrying about the overfitting risk of postprocessing.</p>",
      "rawMarkdown": "Thanks. I am worrying about the overfitting risk of postprocessing.",
      "votes": null
    },
    {
      "id": "658908",
      "postDate": "10/26/2019 18:10:19",
      "content": "<p>Something <em>Steely</em> in that phrase 😜 </p>",
      "rawMarkdown": "Something *Steely* in that phrase 😜",
      "votes": null
    },
    {
      "id": "658934",
      "postDate": "10/26/2019 19:13:48",
      "content": "<p>we're overfitting too much in this competition because the dataset is too noisy and mask size is very low compared to training image size :(\nhaven't face such issue before.</p>",
      "rawMarkdown": "we're overfitting too much in this competition because the dataset is too noisy and mask size is very low compared to training image size :(\nhaven't face such issue before.",
      "votes": null
    },
    {
      "id": "659014",
      "postDate": "10/26/2019 23:40:53",
      "content": "<p>In my understanding on this competition, larger model seems won't help too much.</p>",
      "rawMarkdown": "In my understanding on this competition, larger model seems won't help too much.",
      "votes": null
    },
    {
      "id": "659232",
      "postDate": "10/27/2019 10:08:46",
      "content": "<p>This <a href=\"https://www.kaggle.com/artgor/segmentation-in-pytorch-using-convenient-tools\">kernel</a> gave me a lot of help as a beginner.</p>",
      "rawMarkdown": "This [kernel](https://www.kaggle.com/artgor/segmentation-in-pytorch-using-convenient-tools) gave me a lot of help as a beginner.",
      "votes": null
    },
    {
      "id": "659317",
      "postDate": "10/27/2019 13:01:39",
      "content": "<p>it seems that many kagglers from the steel competition are moving their code to here 😄 </p>",
      "rawMarkdown": "it seems that many kagglers from the steel competition are moving their code to here 😄",
      "votes": null
    },
    {
      "id": "659408",
      "postDate": "10/27/2019 15:57:25",
      "content": "<p>Hi, is it a bad idea?</p>",
      "rawMarkdown": "Hi, is it a bad idea?",
      "votes": null
    },
    {
      "id": "659414",
      "postDate": "10/27/2019 16:13:32",
      "content": "<p>Unfortunately I cannot make BCEDICE loss work with my own pipeline...</p>",
      "rawMarkdown": "Unfortunately I cannot make BCEDICE loss work with my own pipeline...",
      "votes": null
    },
    {
      "id": "659415",
      "postDate": "10/27/2019 16:14:17",
      "content": "<p>You basically only need to change the file directory....</p>",
      "rawMarkdown": "You basically only need to change the file directory....",
      "votes": null
    },
    {
      "id": "659703",
      "postDate": "10/28/2019 05:58:59",
      "content": "<p>Once bitten, twice shy？</p>",
      "rawMarkdown": "Once bitten, twice shy？",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 658632,
      "author_name": "mobassir",
      "author_url": "",
      "post_date": "10/26/2019 09:32:39",
      "content": "<p><a href=\"https://www.kaggle.com/c/understanding_cloud_organization/discussion/110650#latest-639010\">https://www.kaggle.com/c/understanding_cloud_organization/discussion/110650#latest-639010</a></p>\n\n<p><a href=\"https://www.kaggle.com/c/understanding_cloud_organization/discussion/104573#latest-635362\">https://www.kaggle.com/c/understanding_cloud_organization/discussion/104573#latest-635362</a></p>\n\n<p><a href=\"https://www.kaggle.com/c/understanding_cloud_organization/discussion/110303#latest-643328\">https://www.kaggle.com/c/understanding_cloud_organization/discussion/110303#latest-643328</a></p>\n\n<p><a href=\"https://www.kaggle.com/c/understanding_cloud_organization/discussion/108306#latest-631556\">https://www.kaggle.com/c/understanding_cloud_organization/discussion/108306#latest-631556</a></p>\n\n<p>but why you think postprocessing is not a good starting point?</p>",
      "votes": null,
      "replies": [
        {
          "id": 658905,
          "author_name": "naivelamb",
          "author_url": "",
          "post_date": "10/26/2019 18:08:28",
          "content": "<p>Thanks. I am worrying about the overfitting risk of postprocessing.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 658908,
          "author_name": "bibek777",
          "author_url": "",
          "post_date": "10/26/2019 18:10:19",
          "content": "<p>Something <em>Steely</em> in that phrase 😜 </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 658934,
          "author_name": "mobassir",
          "author_url": "",
          "post_date": "10/26/2019 19:13:48",
          "content": "<p>we're overfitting too much in this competition because the dataset is too noisy and mask size is very low compared to training image size :(\nhaven't face such issue before.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 659703,
          "author_name": "hesene",
          "author_url": "",
          "post_date": "10/28/2019 05:58:59",
          "content": "<p>Once bitten, twice shy？</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 659014,
      "author_name": "xiejialun",
      "author_url": "",
      "post_date": "10/26/2019 23:40:53",
      "content": "<p>In my understanding on this competition, larger model seems won't help too much.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 659232,
      "author_name": "wochidadonggua",
      "author_url": "",
      "post_date": "10/27/2019 10:08:46",
      "content": "<p>This <a href=\"https://www.kaggle.com/artgor/segmentation-in-pytorch-using-convenient-tools\">kernel</a> gave me a lot of help as a beginner.</p>",
      "votes": null,
      "replies": [
        {
          "id": 659414,
          "author_name": "naivelamb",
          "author_url": "",
          "post_date": "10/27/2019 16:13:32",
          "content": "<p>Unfortunately I cannot make BCEDICE loss work with my own pipeline...</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 659317,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "10/27/2019 13:01:39",
      "content": "<p>it seems that many kagglers from the steel competition are moving their code to here 😄 </p>",
      "votes": null,
      "replies": [
        {
          "id": 659408,
          "author_name": "mobassir",
          "author_url": "",
          "post_date": "10/27/2019 15:57:25",
          "content": "<p>Hi, is it a bad idea?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 659415,
          "author_name": "naivelamb",
          "author_url": "",
          "post_date": "10/27/2019 16:14:17",
          "content": "<p>You basically only need to change the file directory....</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "658386": "Dear all,\n\nCould you have any suggestions on where to start reading into this competition? Lots of discussions are about postprocessing. Not sure if that's a good starting point.",
    "658632": "https://www.kaggle.com/c/understanding_cloud_organization/discussion/110650#latest-639010\n\nhttps://www.kaggle.com/c/understanding_cloud_organization/discussion/104573#latest-635362\n\nhttps://www.kaggle.com/c/understanding_cloud_organization/discussion/110303#latest-643328\n\nhttps://www.kaggle.com/c/understanding_cloud_organization/discussion/108306#latest-631556\n\nbut why you think postprocessing is not a good starting point?",
    "658905": "Thanks. I am worrying about the overfitting risk of postprocessing.",
    "658908": "Something *Steely* in that phrase 😜",
    "658934": "we're overfitting too much in this competition because the dataset is too noisy and mask size is very low compared to training image size :(\nhaven't face such issue before.",
    "659014": "In my understanding on this competition, larger model seems won't help too much.",
    "659232": "This [kernel](https://www.kaggle.com/artgor/segmentation-in-pytorch-using-convenient-tools) gave me a lot of help as a beginner.",
    "659317": "it seems that many kagglers from the steel competition are moving their code to here 😄",
    "659408": "Hi, is it a bad idea?",
    "659414": "Unfortunately I cannot make BCEDICE loss work with my own pipeline...",
    "659415": "You basically only need to change the file directory....",
    "659703": "Once bitten, twice shy？"
  },
  "source": "meta"
}