{
  "id": 10012,
  "title": "Compression Ratio Benchmark",
  "url": "/competitions/seizure-detection/discussion/10012",
  "author_name": "",
  "post_date": "2014-08-11T22:40:10.487Z",
  "votes": null,
  "comment_count": 3,
  "views": 1470,
  "content": "<p>On the leaderboard, there is an entry stating &quot;Compression Ratio Benchmark&quot;. I don't see what purpose this entry has, does anyone know what can be learnt from this?</p>\n<p>Kind regards</p>",
  "messages": [
    {
      "id": "51890",
      "postDate": "08/11/2014 22:40:10",
      "content": "<p>On the leaderboard, there is an entry stating &quot;Compression Ratio Benchmark&quot;. I don't see what purpose this entry has, does anyone know what can be learnt from this?</p>\n<p>Kind regards</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "51895",
      "postDate": "08/12/2014 01:14:32",
      "content": "<p>http://www.kaggle.com/wiki/Metrics does not mention Compression Ratio Benchmark (CRB). I guess that Kaggle has defined this benchmark based on the CSV file size and Kaggle may not consider submissions with public-leaderboard scores below CRB score for their private scoring. Thanks.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "51901",
      "postDate": "08/12/2014 05:26:46",
      "content": "<p>What can be learned is that the ability to &quot;compress&quot; seizures vs non-seizures is not the same.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "51919",
      "postDate": "08/12/2014 18:51:43",
      "content": "<p>With a compressor you can estimate complexity.</p>\n<p>A compressor does well (gets a high compression ratio) on files that are simple, have repetitive, predictable structure.</p>\n<p>A compressor does poorly (little compression) on files that are complex, with, chaotic, seemingly random structures.</p>\n<p>One would intuitively say that seizures show readings which are&nbsp;complex, chaotic, unpredictable, bizarre. So in general: this complex&nbsp;seizure data would be harder to compress than well-structured, easy-going non-seizure activity -- compression ratio may have a predictive value. The compression ratio benchmark tests this (or a similar) hypothesis.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 51895,
      "author_name": "lalitapatel",
      "author_url": "",
      "post_date": "08/12/2014 01:14:32",
      "content": "<p>http://www.kaggle.com/wiki/Metrics does not mention Compression Ratio Benchmark (CRB). I guess that Kaggle has defined this benchmark based on the CSV file size and Kaggle may not consider submissions with public-leaderboard scores below CRB score for their private scoring. Thanks.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 51901,
      "author_name": "drewabbot",
      "author_url": "",
      "post_date": "08/12/2014 05:26:46",
      "content": "<p>What can be learned is that the ability to &quot;compress&quot; seizures vs non-seizures is not the same.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 51919,
      "author_name": "triskelion",
      "author_url": "",
      "post_date": "08/12/2014 18:51:43",
      "content": "<p>With a compressor you can estimate complexity.</p>\n<p>A compressor does well (gets a high compression ratio) on files that are simple, have repetitive, predictable structure.</p>\n<p>A compressor does poorly (little compression) on files that are complex, with, chaotic, seemingly random structures.</p>\n<p>One would intuitively say that seizures show readings which are&nbsp;complex, chaotic, unpredictable, bizarre. So in general: this complex&nbsp;seizure data would be harder to compress than well-structured, easy-going non-seizure activity -- compression ratio may have a predictive value. The compression ratio benchmark tests this (or a similar) hypothesis.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "51890": "",
    "51895": "",
    "51901": "",
    "51919": ""
  },
  "source": "meta"
}