{
  "id": 115616,
  "title": "Small Shake-up?",
  "url": "/competitions/rsna-intracranial-hemorrhage-detection/discussion/115616",
  "author_name": "Duc Nguyen",
  "post_date": "2019-11-04T05:11:38.467000",
  "votes": 11,
  "comment_count": 6,
  "views": 0,
  "content": "<p>During my experiments, the variance is around 0.0015 -&gt; there will be a small shake-up in the top20. How do you think?</p>",
  "messages": [
    {
      "id": 664707,
      "postDate": "2019-11-04T05:11:38.467Z",
      "content": "<p>During my experiments, the variance is around 0.0015 -&gt; there will be a small shake-up in the top20. How do you think?</p>",
      "rawMarkdown": "During my experiments, the variance is around 0.0015 -&gt; there will be a small shake-up in the top20. How do you think?",
      "votes": 11
    },
    {
      "id": 664762,
      "postDate": "2019-11-04T07:30:46.523Z",
      "content": "<p>Emm. Hope that I'm not overfitting :)</p>",
      "rawMarkdown": "Emm. Hope that I'm not overfitting :)",
      "votes": 8
    },
    {
      "id": 664757,
      "postDate": "2019-11-04T07:14:01.083Z",
      "content": "<p>it is either  \"yes\" or \"no\", depending on the methods.</p>\n\n<p>i find that single image has more variance than multiple image prediction</p>",
      "rawMarkdown": "it is either  \"yes\" or \"no\", depending on the methods.\n\ni find that single image has more variance than multiple image prediction",
      "votes": 2
    },
    {
      "id": 664751,
      "postDate": "2019-11-04T07:04:53.620Z",
      "content": "<p>It's a bit hard to predict with retraining in the pipeline I think (it can go beautifully up or catastrophically down) </p>",
      "rawMarkdown": "It's a bit hard to predict with retraining in the pipeline I think (it can go beautifully up or catastrophically down) "
    },
    {
      "id": 664733,
      "postDate": "2019-11-04T06:30:49.990Z",
      "content": "<p>Maybe you are correct. Let's see the data composition of stage 2. :-)</p>",
      "rawMarkdown": "Maybe you are correct. Let's see the data composition of stage 2. :-)"
    },
    {
      "id": 665145,
      "postDate": "2019-11-04T17:13:43.340Z",
      "rawMarkdown": "",
      "isDeleted": true
    },
    {
      "id": 664990,
      "postDate": "2019-11-04T13:52:16.943Z",
      "content": "<p>thanks</p>",
      "rawMarkdown": "thanks"
    }
  ],
  "comments": [
    {
      "id": 664762,
      "author_name": "SeuTao",
      "author_url": "",
      "post_date": "2019-11-04T07:30:46.523000",
      "content": "<p>Emm. Hope that I'm not overfitting :)</p>",
      "votes": 8,
      "replies": []
    },
    {
      "id": 664757,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "2019-11-04T07:14:01.083000",
      "content": "<p>it is either  \"yes\" or \"no\", depending on the methods.</p>\n\n<p>i find that single image has more variance than multiple image prediction</p>",
      "votes": 2,
      "replies": []
    },
    {
      "id": 664751,
      "author_name": "yukiya",
      "author_url": "",
      "post_date": "2019-11-04T07:04:53.620000",
      "content": "<p>It's a bit hard to predict with retraining in the pipeline I think (it can go beautifully up or catastrophically down) </p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 664733,
      "author_name": "FGPC",
      "author_url": "",
      "post_date": "2019-11-04T06:30:49.990000",
      "content": "<p>Maybe you are correct. Let's see the data composition of stage 2. :-)</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 665145,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-11-04T17:13:43.340000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 664990,
      "author_name": "zhaolie",
      "author_url": "",
      "post_date": "2019-11-04T13:52:16.943000",
      "content": "<p>thanks</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "664707": "During my experiments, the variance is around 0.0015 -&gt; there will be a small shake-up in the top20. How do you think?",
    "664762": "Emm. Hope that I'm not overfitting :)",
    "664757": "it is either  \"yes\" or \"no\", depending on the methods.\n\ni find that single image has more variance than multiple image prediction",
    "664751": "It's a bit hard to predict with retraining in the pipeline I think (it can go beautifully up or catastrophically down) ",
    "664733": "Maybe you are correct. Let's see the data composition of stage 2. :-)",
    "665145": "",
    "664990": "thanks"
  }
}