{
  "id": 183457,
  "title": "Image-level weighted log loss",
  "url": "/competitions/rsna-str-pulmonary-embolism-detection/discussion/183457",
  "author_name": "Abdur Rehman",
  "post_date": "2020-09-16T18:00:34.837000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Image-level log loss given by kaggle:</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F115173%2F190abb127749ab99082a1b0454205b48%2Fql_c2a79726b7d3b29ce07fc58ae568c6e8_l3.png?generation=1599680376511361&amp;alt=media\" alt=\"\"></p>\n<p>where q_i = m_i/n_i is the proportion of positive images in exam i.</p>\n<p>Does this mean we will need an image id and proportion of positive images in the study associated with that image while calculating the loss function? Can someone please clarify this ?</p>",
  "messages": [
    {
      "id": 1013482,
      "postDate": "2020-09-16T18:00:34.837Z",
      "content": "<p>Image-level log loss given by kaggle:</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F115173%2F190abb127749ab99082a1b0454205b48%2Fql_c2a79726b7d3b29ce07fc58ae568c6e8_l3.png?generation=1599680376511361&amp;alt=media\" alt=\"\"></p>\n<p>where q_i = m_i/n_i is the proportion of positive images in exam i.</p>\n<p>Does this mean we will need an image id and proportion of positive images in the study associated with that image while calculating the loss function? Can someone please clarify this ?</p>",
      "rawMarkdown": "Image-level log loss given by kaggle:\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F115173%2F190abb127749ab99082a1b0454205b48%2Fql_c2a79726b7d3b29ce07fc58ae568c6e8_l3.png?generation=1599680376511361&alt=media)\n\nwhere q_i = m_i/n_i is the proportion of positive images in exam i.\n\nDoes this mean we will need an image id and proportion of positive images in the study associated with that image while calculating the loss function? Can someone please clarify this ?"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1013482": "Image-level log loss given by kaggle:\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F115173%2F190abb127749ab99082a1b0454205b48%2Fql_c2a79726b7d3b29ce07fc58ae568c6e8_l3.png?generation=1599680376511361&alt=media)\n\nwhere q_i = m_i/n_i is the proportion of positive images in exam i.\n\nDoes this mean we will need an image id and proportion of positive images in the study associated with that image while calculating the loss function? Can someone please clarify this ?"
  }
}