{
  "id": 243935,
  "title": "anyone try human auc score?",
  "url": "/competitions/seti-breakthrough-listen/discussion/243935",
  "author_name": "assign",
  "post_date": "2021-06-04T14:06:11.246000",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I train on last 3000 data and<br>\ntest on first 880 data and get 0.77auc …..<br>\ntrain time : 3.6H, epochs 2<br>\ntest time : 1.3H</p>\n<p>no augmentation, no Kfold<br>\nwithout pseudo labeling : 0.77 auc , train auc = 0.81<br>\nwith pseudo labeling : 0.64 auc  (\"without pseudo labeling\" model is reused, different 880 data)</p>\n<p>I found there are lots of false cases with single evident line<br>\nI can't understand AI </p>",
  "messages": [
    {
      "id": 1335916,
      "postDate": "2021-06-04T14:06:11.247Z",
      "content": "<p>I train on last 3000 data and<br>\ntest on first 880 data and get 0.77auc …..<br>\ntrain time : 3.6H, epochs 2<br>\ntest time : 1.3H</p>\n<p>no augmentation, no Kfold<br>\nwithout pseudo labeling : 0.77 auc , train auc = 0.81<br>\nwith pseudo labeling : 0.64 auc  (\"without pseudo labeling\" model is reused, different 880 data)</p>\n<p>I found there are lots of false cases with single evident line<br>\nI can't understand AI </p>",
      "rawMarkdown": "I train on last 3000 data and\ntest on first 880 data and get 0.77auc .....\ntrain time : 3.6H, epochs 2\ntest time : 1.3H\n\nno augmentation, no Kfold\nwithout pseudo labeling : 0.77 auc , train auc = 0.81\nwith pseudo labeling : 0.64 auc  (\"without pseudo labeling\" model is reused, different 880 data)\n\nI found there are lots of false cases with single evident line\nI can't understand AI ",
      "votes": 2
    },
    {
      "id": 1355556,
      "postDate": "2021-06-18T11:12:19.640Z",
      "content": "<p>You were on to something, huh?</p>\n<p>In the olden days, the host would provide a submission with their best result as a good benchmark; that would have helped here from the start.</p>",
      "rawMarkdown": "You were on to something, huh?\n\nIn the olden days, the host would provide a submission with their best result as a good benchmark; that would have helped here from the start."
    }
  ],
  "comments": [
    {
      "id": 1355556,
      "author_name": "Tord Malmgren",
      "author_url": "",
      "post_date": "2021-06-18T11:12:19.640000",
      "content": "<p>You were on to something, huh?</p>\n<p>In the olden days, the host would provide a submission with their best result as a good benchmark; that would have helped here from the start.</p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1335916": "I train on last 3000 data and\ntest on first 880 data and get 0.77auc .....\ntrain time : 3.6H, epochs 2\ntest time : 1.3H\n\nno augmentation, no Kfold\nwithout pseudo labeling : 0.77 auc , train auc = 0.81\nwith pseudo labeling : 0.64 auc  (\"without pseudo labeling\" model is reused, different 880 data)\n\nI found there are lots of false cases with single evident line\nI can't understand AI ",
    "1355556": "You were on to something, huh?\n\nIn the olden days, the host would provide a submission with their best result as a good benchmark; that would have helped here from the start."
  }
}