{
  "id": 7970,
  "title": "Is clustering discouraged?",
  "url": "/competitions/decoding-the-human-brain/discussion/7970",
  "author_name": "",
  "post_date": "2014-05-02T10:43:04.140Z",
  "votes": null,
  "comment_count": 1,
  "views": 844,
  "content": "<p>In theory, one could employ unsupervised techniques to cluster the data of a subject from the unlabeled testing set in a manner independent of the data from the subjects in the labeled training set. This approach increases in feasibility when considering only two classes. However, this does not seem to be what the Competition Administrator desires based on the description of the competition and the linked paper. It appears that the objective of the competition is to investigate methods for&nbsp;<em>transferring</em>&nbsp;a collection of decoding models developed on multiple subjects to new subjects. Is this a correct assessment? If so, would the Competition Administrator be less than satisfied with a clustering solution?</p>",
  "messages": [
    {
      "id": "43506",
      "postDate": "05/02/2014 10:43:04",
      "content": "<p>In theory, one could employ unsupervised techniques to cluster the data of a subject from the unlabeled testing set in a manner independent of the data from the subjects in the labeled training set. This approach increases in feasibility when considering only two classes. However, this does not seem to be what the Competition Administrator desires based on the description of the competition and the linked paper. It appears that the objective of the competition is to investigate methods for&nbsp;<em>transferring</em>&nbsp;a collection of decoding models developed on multiple subjects to new subjects. Is this a correct assessment? If so, would the Competition Administrator be less than satisfied with a clustering solution?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "43516",
      "postDate": "05/02/2014 15:01:39",
      "content": "<p>Hi,</p>\n<p>The description of the competition and the&nbsp;<a href=\"http://arxiv.org/abs/1404.4175\">linked paper</a>&nbsp;that you mention,&nbsp;describe only some of the ways to approach the problem of the competition. Every solution that proves to be effective is certainly welcome. If clustering or other unsupervised methods are helpful to predict the stimuli of the test subjects, we invite the participants to adopt them.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 43516,
      "author_name": "emanuele",
      "author_url": "",
      "post_date": "05/02/2014 15:01:39",
      "content": "<p>Hi,</p>\n<p>The description of the competition and the&nbsp;<a href=\"http://arxiv.org/abs/1404.4175\">linked paper</a>&nbsp;that you mention,&nbsp;describe only some of the ways to approach the problem of the competition. Every solution that proves to be effective is certainly welcome. If clustering or other unsupervised methods are helpful to predict the stimuli of the test subjects, we invite the participants to adopt them.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "43506": "",
    "43516": ""
  },
  "source": "meta"
}