{
  "id": 5789,
  "title": "Ideas on downsampling FFT?",
  "url": "/competitions/belkin-energy-disaggregation-competition/discussion/5789",
  "author_name": "gallamine",
  "post_date": "2013-09-17T15:54:43.803000",
  "votes": 0,
  "comment_count": 0,
  "views": 815,
  "content": "<p>Does anyone have any suggestions for reducing the size of the FFT? At first I considered using an autoencoder, but I'm having trouble with the vectors being very similar. Another thought would be to do N-peak picking, but I'm also concerned that systemic in-house features will be present in every example.</p>\n<p>I know there's issues with this competition, but I'm trying to move past them and at least make a concerted effort to get past the all-zero benchmark.</p>",
  "messages": [
    {
      "id": 31052,
      "postDate": "2013-09-17T15:54:43.803Z",
      "content": "<p>Does anyone have any suggestions for reducing the size of the FFT? At first I considered using an autoencoder, but I'm having trouble with the vectors being very similar. Another thought would be to do N-peak picking, but I'm also concerned that systemic in-house features will be present in every example.</p>\n<p>I know there's issues with this competition, but I'm trying to move past them and at least make a concerted effort to get past the all-zero benchmark.</p>"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "31052": ""
  }
}