{
  "id": 162779,
  "title": "Why not convert audio data into tabular data for prediction",
  "url": "/competitions/birdsong-recognition/discussion/162779",
  "author_name": "",
  "post_date": "2020-06-30T03:00:29.024285Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I see that the vast majority of public kernels use neural networks as the basic model like lgb, xgb , but nobody uses the tree model. I feel very strange.</p>\n\n<p>This is normal?</p>",
  "messages": [
    {
      "id": "907546",
      "postDate": "06/30/2020 03:00:29",
      "content": "<p>I see that the vast majority of public kernels use neural networks as the basic model like lgb, xgb , but nobody uses the tree model. I feel very strange.</p>\n\n<p>This is normal?</p>",
      "rawMarkdown": "I see that the vast majority of public kernels use neural networks as the basic model like lgb, xgb , but nobody uses the tree model. I feel very strange.\n\nThis is normal?",
      "votes": null
    },
    {
      "id": "907890",
      "postDate": "06/30/2020 08:15:00",
      "content": "<p>In 2013 I was able to win a bird song classification challenge with spectrogram template matching and random forests. That time we had only 500 audio files for training and we did not have comfortable deep learning tools. I don't think that tree based methods would be competitive for this challenge.</p>\n\n<p>Anyway feel free to try any approach!</p>",
      "rawMarkdown": "In 2013 I was able to win a bird song classification challenge with spectrogram template matching and random forests. That time we had only 500 audio files for training and we did not have comfortable deep learning tools. I don't think that tree based methods would be competitive for this challenge.\n\nAnyway feel free to try any approach!",
      "votes": null
    },
    {
      "id": "910523",
      "postDate": "07/01/2020 07:36:40",
      "content": "<p>Thank you for your reply, did you use the notebook of Random Forest before?</p>",
      "rawMarkdown": "Thank you for your reply, did you use the notebook of Random Forest before?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 907890,
      "author_name": "gaborfodor",
      "author_url": "",
      "post_date": "06/30/2020 08:15:00",
      "content": "<p>In 2013 I was able to win a bird song classification challenge with spectrogram template matching and random forests. That time we had only 500 audio files for training and we did not have comfortable deep learning tools. I don't think that tree based methods would be competitive for this challenge.</p>\n\n<p>Anyway feel free to try any approach!</p>",
      "votes": null,
      "replies": [
        {
          "id": 910523,
          "author_name": "xiaowangiiiii",
          "author_url": "",
          "post_date": "07/01/2020 07:36:40",
          "content": "<p>Thank you for your reply, did you use the notebook of Random Forest before?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "907546": "I see that the vast majority of public kernels use neural networks as the basic model like lgb, xgb , but nobody uses the tree model. I feel very strange.\n\nThis is normal?",
    "907890": "In 2013 I was able to win a bird song classification challenge with spectrogram template matching and random forests. That time we had only 500 audio files for training and we did not have comfortable deep learning tools. I don't think that tree based methods would be competitive for this challenge.\n\nAnyway feel free to try any approach!",
    "910523": "Thank you for your reply, did you use the notebook of Random Forest before?"
  },
  "source": "meta"
}