{
  "id": 339788,
  "title": "Scope of competition: Unsupervised algorithms?",
  "url": "/competitions/energy-anomaly-detection/discussion/339788",
  "author_name": "",
  "post_date": "2022-07-26T11:34:17.045361500Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I see from the public notebooks many supervised approaches, ie. treat it as an imbalanced binary classification problem. </p>\n<p>Having in mind that in many real applications such labeled data don't exist in general, or are very very limited. <br>\nIt would be very interesting to try unsupervised methods for anomaly detection and use the given annotated samples for evaluation only.   </p>\n<p>Any ideas?  Anyone tried something similar? </p>",
  "messages": [
    {
      "id": "1871619",
      "postDate": "07/26/2022 11:34:17",
      "content": "<p>I see from the public notebooks many supervised approaches, ie. treat it as an imbalanced binary classification problem. </p>\n<p>Having in mind that in many real applications such labeled data don't exist in general, or are very very limited. <br>\nIt would be very interesting to try unsupervised methods for anomaly detection and use the given annotated samples for evaluation only.   </p>\n<p>Any ideas?  Anyone tried something similar? </p>",
      "rawMarkdown": "I see from the public notebooks many supervised approaches, ie. treat it as an imbalanced binary classification problem. \n\nHaving in mind that in many real applications such labeled data don't exist in general, or are very very limited. \nIt would be very interesting to try unsupervised methods for anomaly detection and use the given annotated samples for evaluation only.   \n\nAny ideas?  Anyone tried something similar?",
      "votes": null
    },
    {
      "id": "1877033",
      "postDate": "07/30/2022 09:30:10",
      "content": "<p>I think it's an interesting idea, but I didn't get any interesting results when I tried reinforcement learning and I couldn't investigate more on the results because of time constraints.</p>",
      "rawMarkdown": "I think it's an interesting idea, but I didn't get any interesting results when I tried reinforcement learning and I couldn't investigate more on the results because of time constraints.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1877033,
      "author_name": "kamdoumloich",
      "author_url": "",
      "post_date": "07/30/2022 09:30:10",
      "content": "<p>I think it's an interesting idea, but I didn't get any interesting results when I tried reinforcement learning and I couldn't investigate more on the results because of time constraints.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1871619": "I see from the public notebooks many supervised approaches, ie. treat it as an imbalanced binary classification problem. \n\nHaving in mind that in many real applications such labeled data don't exist in general, or are very very limited. \nIt would be very interesting to try unsupervised methods for anomaly detection and use the given annotated samples for evaluation only.   \n\nAny ideas?  Anyone tried something similar?",
    "1877033": "I think it's an interesting idea, but I didn't get any interesting results when I tried reinforcement learning and I couldn't investigate more on the results because of time constraints."
  },
  "source": "meta"
}