{
  "id": 11457,
  "title": "Image Sampling Method",
  "url": "/competitions/datasciencebowl/discussion/11457",
  "author_name": "Shariq Hashme",
  "post_date": "2015-01-09T00:03:06.560000",
  "votes": 2,
  "comment_count": 0,
  "views": 658,
  "content": "<p>The competition rules mention &quot;no external data&quot;.<br><br>How were the images sampled? Is it fair to assume these were sampled randomly from the&nbsp;larger dataset we will be tested against in the end?<br><br>Asking because it seems like the distribution of samples within classes is very skewed,&nbsp;so it might&nbsp;make sense to somehow weight predictions&nbsp;based on the expected frequency of that class showing up.</p>",
  "messages": [
    {
      "id": 62474,
      "postDate": "2015-01-09T00:03:06.560Z",
      "content": "<p>The competition rules mention &quot;no external data&quot;.<br><br>How were the images sampled? Is it fair to assume these were sampled randomly from the&nbsp;larger dataset we will be tested against in the end?<br><br>Asking because it seems like the distribution of samples within classes is very skewed,&nbsp;so it might&nbsp;make sense to somehow weight predictions&nbsp;based on the expected frequency of that class showing up.</p>",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "62474": ""
  }
}