{
  "id": 45653,
  "title": "Some files make cost down to 0",
  "url": "/competitions/tensorflow-speech-recognition-challenge/discussion/45653",
  "author_name": "IsaacSim",
  "post_date": "2017-12-14T08:01:10.630000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi, everyone.</p>\n\n<p>I made a simple deep learning neural network for it,\nwhich it was working for other tensorflow image set (mnist cifar 10)</p>\n\n<p>But while I was working on this competition some(actually many) files causes the cost down to 0 with the same model I used above. So I tracked back to listen to the wav file, and IT WAS OKAY. I am sure this is pretty much with data or data transfer to vectorized form.</p>\n\n<p>Obviously I used the data set with 16000 data pieces in it, and padded some that are short and cut some that are long(like noisy folder)</p>\n\n<p>Does anyone have same problem with me? and how did you guys manage ?</p>",
  "messages": [
    {
      "id": 257422,
      "postDate": "2017-12-14T08:01:10.630Z",
      "content": "<p>Hi, everyone.</p>\n\n<p>I made a simple deep learning neural network for it,\nwhich it was working for other tensorflow image set (mnist cifar 10)</p>\n\n<p>But while I was working on this competition some(actually many) files causes the cost down to 0 with the same model I used above. So I tracked back to listen to the wav file, and IT WAS OKAY. I am sure this is pretty much with data or data transfer to vectorized form.</p>\n\n<p>Obviously I used the data set with 16000 data pieces in it, and padded some that are short and cut some that are long(like noisy folder)</p>\n\n<p>Does anyone have same problem with me? and how did you guys manage ?</p>",
      "rawMarkdown": "Hi, everyone.\n\nI made a simple deep learning neural network for it,\nwhich it was working for other tensorflow image set (mnist cifar 10)\n\nBut while I was working on this competition some(actually many) files causes the cost down to 0 with the same model I used above. So I tracked back to listen to the wav file, and IT WAS OKAY. I am sure this is pretty much with data or data transfer to vectorized form.\n\nObviously I used the data set with 16000 data pieces in it, and padded some that are short and cut some that are long(like noisy folder)\n\nDoes anyone have same problem with me? and how did you guys manage ?"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "257422": "Hi, everyone.\n\nI made a simple deep learning neural network for it,\nwhich it was working for other tensorflow image set (mnist cifar 10)\n\nBut while I was working on this competition some(actually many) files causes the cost down to 0 with the same model I used above. So I tracked back to listen to the wav file, and IT WAS OKAY. I am sure this is pretty much with data or data transfer to vectorized form.\n\nObviously I used the data set with 16000 data pieces in it, and padded some that are short and cut some that are long(like noisy folder)\n\nDoes anyone have same problem with me? and how did you guys manage ?"
  }
}