{
  "id": 572551,
  "title": "Possibly wrong definition of false negative in this competition?",
  "url": "/competitions/byu-locating-bacterial-flagellar-motors-2025/discussion/572551",
  "author_name": "",
  "post_date": "2025-04-10T04:00:49.927115100Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi,</p>\n<p>The following definitions of TP and FN are given in the overview section of this competition</p>\n<ul>\n<li>True Positive (TP): If |y−y¯|2≤τ, the prediction is within threshold.</li>\n<li>False Negative (FN): If |y−y¯|2&gt;τ, the prediction is outside of threshold.</li>\n</ul>\n<p>where  τ=1000 Angstroms.</p>\n<p>However, isn't the false negative defined here actually a false positive? There is a prediction so it's positive, but its false (outside the tau threshold) so it's a false positive. </p>\n<p>Wouldn't a false negative occur if given a tomogram with flagellar motors somewhere in it, the model prediction would be  -1, -1, -1 (no motors found)?</p>",
  "messages": [
    {
      "id": "3175354",
      "postDate": "04/10/2025 04:00:49",
      "content": "<p>Hi,</p>\n<p>The following definitions of TP and FN are given in the overview section of this competition</p>\n<ul>\n<li>True Positive (TP): If |y−y¯|2≤τ, the prediction is within threshold.</li>\n<li>False Negative (FN): If |y−y¯|2&gt;τ, the prediction is outside of threshold.</li>\n</ul>\n<p>where  τ=1000 Angstroms.</p>\n<p>However, isn't the false negative defined here actually a false positive? There is a prediction so it's positive, but its false (outside the tau threshold) so it's a false positive. </p>\n<p>Wouldn't a false negative occur if given a tomogram with flagellar motors somewhere in it, the model prediction would be  -1, -1, -1 (no motors found)?</p>",
      "rawMarkdown": "Hi,\n\nThe following definitions of TP and FN are given in the overview section of this competition\n\n- True Positive (TP): If |y−y¯|2≤τ, the prediction is within threshold.\n- False Negative (FN): If |y−y¯|2>τ, the prediction is outside of threshold.\n\nwhere  τ=1000 Angstroms.\n\nHowever, isn't the false negative defined here actually a false positive? There is a prediction so it's positive, but its false (outside the tau threshold) so it's a false positive. \n\nWouldn't a false negative occur if given a tomogram with flagellar motors somewhere in it, the model prediction would be  -1, -1, -1 (no motors found)?",
      "votes": null
    },
    {
      "id": "3175385",
      "postDate": "04/10/2025 04:48:24",
      "content": "<blockquote>\n  <p>Wouldn't a false negative occur if given a tomogram with flagellar motors somewhere in it, the model prediction would be -1, -1, -1 (no motors found)?</p>\n</blockquote>\n<p>These both would be false negatives according to the metric. If you predict outside the threshold, the prediction is as effective as no prediction at all, and so we handle them the same as the standard false negative case. A false positive would be predicting that there is a motor on a tomogram without any motors. </p>\n<p>With this definition, precision becomes a task of not misidentifying motors, and recall a task of locating their position. And since recall is weighted higher in this competition, it serves our purposes. </p>\n<p>Granted, you could easily consider these false negatives to be false positives. It's a matter of opinion. We designed the metric around these definitions. I'm sorry if it's confusing, I hope this helps clarify things. </p>",
      "rawMarkdown": ">Wouldn't a false negative occur if given a tomogram with flagellar motors somewhere in it, the model prediction would be -1, -1, -1 (no motors found)?\n\nThese both would be false negatives according to the metric. If you predict outside the threshold, the prediction is as effective as no prediction at all, and so we handle them the same as the standard false negative case. A false positive would be predicting that there is a motor on a tomogram without any motors. \n\nWith this definition, precision becomes a task of not misidentifying motors, and recall a task of locating their position. And since recall is weighted higher in this competition, it serves our purposes. \n\nGranted, you could easily consider these false negatives to be false positives. It's a matter of opinion. We designed the metric around these definitions. I'm sorry if it's confusing, I hope this helps clarify things.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3175385,
      "author_name": "andrewjdarley",
      "author_url": "",
      "post_date": "04/10/2025 04:48:24",
      "content": "<blockquote>\n  <p>Wouldn't a false negative occur if given a tomogram with flagellar motors somewhere in it, the model prediction would be -1, -1, -1 (no motors found)?</p>\n</blockquote>\n<p>These both would be false negatives according to the metric. If you predict outside the threshold, the prediction is as effective as no prediction at all, and so we handle them the same as the standard false negative case. A false positive would be predicting that there is a motor on a tomogram without any motors. </p>\n<p>With this definition, precision becomes a task of not misidentifying motors, and recall a task of locating their position. And since recall is weighted higher in this competition, it serves our purposes. </p>\n<p>Granted, you could easily consider these false negatives to be false positives. It's a matter of opinion. We designed the metric around these definitions. I'm sorry if it's confusing, I hope this helps clarify things. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3175354": "Hi,\n\nThe following definitions of TP and FN are given in the overview section of this competition\n\n- True Positive (TP): If |y−y¯|2≤τ, the prediction is within threshold.\n- False Negative (FN): If |y−y¯|2>τ, the prediction is outside of threshold.\n\nwhere  τ=1000 Angstroms.\n\nHowever, isn't the false negative defined here actually a false positive? There is a prediction so it's positive, but its false (outside the tau threshold) so it's a false positive. \n\nWouldn't a false negative occur if given a tomogram with flagellar motors somewhere in it, the model prediction would be  -1, -1, -1 (no motors found)?",
    "3175385": ">Wouldn't a false negative occur if given a tomogram with flagellar motors somewhere in it, the model prediction would be -1, -1, -1 (no motors found)?\n\nThese both would be false negatives according to the metric. If you predict outside the threshold, the prediction is as effective as no prediction at all, and so we handle them the same as the standard false negative case. A false positive would be predicting that there is a motor on a tomogram without any motors. \n\nWith this definition, precision becomes a task of not misidentifying motors, and recall a task of locating their position. And since recall is weighted higher in this competition, it serves our purposes. \n\nGranted, you could easily consider these false negatives to be false positives. It's a matter of opinion. We designed the metric around these definitions. I'm sorry if it's confusing, I hope this helps clarify things."
  },
  "source": "meta"
}