{
  "id": 196945,
  "title": "LSTM  for INGV  - Volcanic Eruption Prediction",
  "url": "/competitions/predict-volcanic-eruptions-ingv-oe/discussion/196945",
  "author_name": "",
  "post_date": "2020-11-13T13:59:39.178609500Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Are the sequences too long ~60k for an LSTM?  If no how to set the parameter of the LSTM? If yes is there any way to deal with that?  I suggest to downsamle the sequence by choosing some random  observations to reduce the length of the observation! ( I don't think it will work well).  <br>\nbest regards :) </p>",
  "messages": [
    {
      "id": "1077352",
      "postDate": "11/13/2020 13:59:39",
      "content": "<p>Are the sequences too long ~60k for an LSTM?  If no how to set the parameter of the LSTM? If yes is there any way to deal with that?  I suggest to downsamle the sequence by choosing some random  observations to reduce the length of the observation! ( I don't think it will work well).  <br>\nbest regards :) </p>",
      "rawMarkdown": "Are the sequences too long ~60k for an LSTM?  If no how to set the parameter of the LSTM? If yes is there any way to deal with that?  I suggest to downsamle the sequence by choosing some random  observations to reduce the length of the observation! ( I don't think it will work well).  \nbest regards :)",
      "votes": null
    },
    {
      "id": "1077453",
      "postDate": "11/13/2020 16:19:17",
      "content": "<p>I tested at the beginning of the comp and it does not work</p>",
      "rawMarkdown": "I tested at the beginning of the comp and it does not work",
      "votes": null
    },
    {
      "id": "1087378",
      "postDate": "11/22/2020 16:46:08",
      "content": "<p>I tried that, but did not get a good result. I published a notebook with the LSTM code.</p>",
      "rawMarkdown": "I tried that, but did not get a good result. I published a notebook with the LSTM code.",
      "votes": null
    },
    {
      "id": "1090300",
      "postDate": "11/25/2020 08:41:21",
      "content": "<p>Try Spectrograms. For me window of 2048 points gave the best result.</p>",
      "rawMarkdown": "Try Spectrograms. For me window of 2048 points gave the best result.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1077453,
      "author_name": "enric1296",
      "author_url": "",
      "post_date": "11/13/2020 16:19:17",
      "content": "<p>I tested at the beginning of the comp and it does not work</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1087378,
      "author_name": "catadanna",
      "author_url": "",
      "post_date": "11/22/2020 16:46:08",
      "content": "<p>I tried that, but did not get a good result. I published a notebook with the LSTM code.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1090300,
      "author_name": "michaelmarkzon",
      "author_url": "",
      "post_date": "11/25/2020 08:41:21",
      "content": "<p>Try Spectrograms. For me window of 2048 points gave the best result.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1077352": "Are the sequences too long ~60k for an LSTM?  If no how to set the parameter of the LSTM? If yes is there any way to deal with that?  I suggest to downsamle the sequence by choosing some random  observations to reduce the length of the observation! ( I don't think it will work well).  \nbest regards :)",
    "1077453": "I tested at the beginning of the comp and it does not work",
    "1087378": "I tried that, but did not get a good result. I published a notebook with the LSTM code.",
    "1090300": "Try Spectrograms. For me window of 2048 points gave the best result."
  },
  "source": "meta"
}