{
  "id": 117616,
  "title": "Probable shakeup?",
  "url": "/competitions/understanding_cloud_organization/discussion/117616",
  "author_name": "",
  "post_date": "2019-11-16T18:26:23.253632500Z",
  "votes": 5,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Because of noisy masks, My submission scores are not very consistant. Slight change in parameters gives huge change in scores. I think private LB will be a disaster and I am not feeling confident about winning a medal here.</p>",
  "messages": [
    {
      "id": "674582",
      "postDate": "11/16/2019 18:26:23",
      "content": "<p>Because of noisy masks, My submission scores are not very consistant. Slight change in parameters gives huge change in scores. I think private LB will be a disaster and I am not feeling confident about winning a medal here.</p>",
      "rawMarkdown": "Because of noisy masks, My submission scores are not very consistant. Slight change in parameters gives huge change in scores. I think private LB will be a disaster and I am not feeling confident about winning a medal here.",
      "votes": null
    },
    {
      "id": "674709",
      "postDate": "11/16/2019 23:53:36",
      "content": "<p>It will depend actually. If you use the enseble output without much processing the score should be fine I guess. </p>",
      "rawMarkdown": "It will depend actually. If you use the enseble output without much processing the score should be fine I guess.",
      "votes": null
    },
    {
      "id": "674753",
      "postDate": "11/17/2019 02:46:54",
      "content": "<p>I believe that fine tuning the post-processing parameters will lead to shake up. IMHO, if you are ensembling models to bring diversity and not fine-tuning minsize and mask threshold you can maintain your rank!</p>",
      "rawMarkdown": "I believe that fine tuning the post-processing parameters will lead to shake up. IMHO, if you are ensembling models to bring diversity and not fine-tuning minsize and mask threshold you can maintain your rank!",
      "votes": null
    },
    {
      "id": "674792",
      "postDate": "11/17/2019 04:44:06",
      "content": "<p>Only maintain <a href=\"/adish333\">@adish333</a>? I am hoping for a jump :-)</p>",
      "rawMarkdown": "Only maintain @adish333? I am hoping for a jump :-)",
      "votes": null
    },
    {
      "id": "674870",
      "postDate": "11/17/2019 07:11:29",
      "content": "<p>Yeah, hope for the best 💯 </p>",
      "rawMarkdown": "Yeah, hope for the best 💯",
      "votes": null
    },
    {
      "id": "676093",
      "postDate": "11/19/2019 00:40:40",
      "content": "<p>i used gridsearch for threshholds from 0.3 to 0.7 in intervals of 0.05 and min size from 7.5k to 17.5k but it didn't end up hurting my rank, in fact i jumped ~100 places, though i did use an ensemble of 4 different models</p>",
      "rawMarkdown": "i used gridsearch for threshholds from 0.3 to 0.7 in intervals of 0.05 and min size from 7.5k to 17.5k but it didn't end up hurting my rank, in fact i jumped ~100 places, though i did use an ensemble of 4 different models",
      "votes": null
    },
    {
      "id": "676137",
      "postDate": "11/19/2019 01:40:15",
      "content": "<p>Congrats <a href=\"/sidhanthholalkere\">@sidhanthholalkere</a>. </p>\n\n<p><a href=\"/adish333\">@adish333</a>, only 15 places up for me so it is more like taking a long deep breath than a jump :-)</p>",
      "rawMarkdown": "Congrats @sidhanthholalkere. \n\n@adish333, only 15 places up for me so it is more like taking a long deep breath than a jump :-)",
      "votes": null
    },
    {
      "id": "676169",
      "postDate": "11/19/2019 02:14:26",
      "content": "<p>Infact I am quite surprised that almost no shakeup at all. I retained almost same rank.</p>",
      "rawMarkdown": "Infact I am quite surprised that almost no shakeup at all. I retained almost same rank.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 674709,
      "author_name": "mykttu",
      "author_url": "",
      "post_date": "11/16/2019 23:53:36",
      "content": "<p>It will depend actually. If you use the enseble output without much processing the score should be fine I guess. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 674753,
      "author_name": "adish333",
      "author_url": "",
      "post_date": "11/17/2019 02:46:54",
      "content": "<p>I believe that fine tuning the post-processing parameters will lead to shake up. IMHO, if you are ensembling models to bring diversity and not fine-tuning minsize and mask threshold you can maintain your rank!</p>",
      "votes": null,
      "replies": [
        {
          "id": 674792,
          "author_name": "sheriytm",
          "author_url": "",
          "post_date": "11/17/2019 04:44:06",
          "content": "<p>Only maintain <a href=\"/adish333\">@adish333</a>? I am hoping for a jump :-)</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 674870,
          "author_name": "adish333",
          "author_url": "",
          "post_date": "11/17/2019 07:11:29",
          "content": "<p>Yeah, hope for the best 💯 </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 676093,
          "author_name": "sidhanthholalkere",
          "author_url": "",
          "post_date": "11/19/2019 00:40:40",
          "content": "<p>i used gridsearch for threshholds from 0.3 to 0.7 in intervals of 0.05 and min size from 7.5k to 17.5k but it didn't end up hurting my rank, in fact i jumped ~100 places, though i did use an ensemble of 4 different models</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 676137,
          "author_name": "sheriytm",
          "author_url": "",
          "post_date": "11/19/2019 01:40:15",
          "content": "<p>Congrats <a href=\"/sidhanthholalkere\">@sidhanthholalkere</a>. </p>\n\n<p><a href=\"/adish333\">@adish333</a>, only 15 places up for me so it is more like taking a long deep breath than a jump :-)</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 676169,
          "author_name": "raghaw",
          "author_url": "",
          "post_date": "11/19/2019 02:14:26",
          "content": "<p>Infact I am quite surprised that almost no shakeup at all. I retained almost same rank.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "674582": "Because of noisy masks, My submission scores are not very consistant. Slight change in parameters gives huge change in scores. I think private LB will be a disaster and I am not feeling confident about winning a medal here.",
    "674709": "It will depend actually. If you use the enseble output without much processing the score should be fine I guess.",
    "674753": "I believe that fine tuning the post-processing parameters will lead to shake up. IMHO, if you are ensembling models to bring diversity and not fine-tuning minsize and mask threshold you can maintain your rank!",
    "674792": "Only maintain @adish333? I am hoping for a jump :-)",
    "674870": "Yeah, hope for the best 💯",
    "676093": "i used gridsearch for threshholds from 0.3 to 0.7 in intervals of 0.05 and min size from 7.5k to 17.5k but it didn't end up hurting my rank, in fact i jumped ~100 places, though i did use an ensemble of 4 different models",
    "676137": "Congrats @sidhanthholalkere. \n\n@adish333, only 15 places up for me so it is more like taking a long deep breath than a jump :-)",
    "676169": "Infact I am quite surprised that almost no shakeup at all. I retained almost same rank."
  },
  "source": "meta"
}