{
  "id": 77718,
  "title": "A Thought Regarding 4 MHz Sampling Rate",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/77718",
  "author_name": "",
  "post_date": "2019-01-16T01:09:06.895688900Z",
  "votes": 4,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Hi all,</p>\n\n<p>The sampling rate is 4 MHz which looks very high to me. \nI don't know if this is a common sampling rate deployed in \nthe real earthquake monitoring task, nor do I know if this is common\nfrequency to 'see' the destructive waves (in lab conditions). </p>\n\n<p>Being high is not bad. Resolution thing needs to be just right.</p>\n\n<p>Supposedly,  those major destructive waves, indeed, reside in some\nlower frequency bands. What we will need is to zoom-in our spectral \ngraph, especially the lower chunk.</p>\n\n<p>That is to require a higher resolution in the frequency domain in whichever\nthe transformation we use e.g. STFT</p>\n\n<p>Just realize this. I'm not sure if this will lead to a better modeling strategy.\nIf anyone knows about this, feel free to comment!</p>",
  "messages": [
    {
      "id": "456522",
      "postDate": "01/16/2019 01:09:06",
      "content": "<p>Hi all,</p>\n\n<p>The sampling rate is 4 MHz which looks very high to me. \nI don't know if this is a common sampling rate deployed in \nthe real earthquake monitoring task, nor do I know if this is common\nfrequency to 'see' the destructive waves (in lab conditions). </p>\n\n<p>Being high is not bad. Resolution thing needs to be just right.</p>\n\n<p>Supposedly,  those major destructive waves, indeed, reside in some\nlower frequency bands. What we will need is to zoom-in our spectral \ngraph, especially the lower chunk.</p>\n\n<p>That is to require a higher resolution in the frequency domain in whichever\nthe transformation we use e.g. STFT</p>\n\n<p>Just realize this. I'm not sure if this will lead to a better modeling strategy.\nIf anyone knows about this, feel free to comment!</p>",
      "rawMarkdown": "Hi all,\n\nThe sampling rate is 4 MHz which looks very high to me. \nI don't know if this is a common sampling rate deployed in \nthe real earthquake monitoring task, nor do I know if this is common\nfrequency to 'see' the destructive waves (in lab conditions). \n\nBeing high is not bad. Resolution thing needs to be just right.\n\nSupposedly,  those major destructive waves, indeed, reside in some\nlower frequency bands. What we will need is to zoom-in our spectral \ngraph, especially the lower chunk.\n\nThat is to require a higher resolution in the frequency domain in whichever\nthe transformation we use e.g. STFT\n\nJust realize this. I'm not sure if this will lead to a better modeling strategy.\nIf anyone knows about this, feel free to comment!",
      "votes": null
    },
    {
      "id": "456570",
      "postDate": "01/16/2019 04:21:43",
      "content": "<p>Real world earthquake monitoring uses much lower sample rates - maybe 20 Hz max. The waves travel hundreds of miles and are subject to plenty of attenuation, especially the high frequencies. In the lab, the distance is a few feet and so the frequencies are much higher. Just what type of waves we are seeing in these laboratory data is a bit unclear to me, especially since I don't know the geometry - probably mostly surface waves. \nAlso, I don't think we are looking here at destructive waves - note that the seismic amplitude at the failure is small - no spikes there. The paper calls them \"slow earthquakes\".</p>",
      "rawMarkdown": "Real world earthquake monitoring uses much lower sample rates - maybe 20 Hz max. The waves travel hundreds of miles and are subject to plenty of attenuation, especially the high frequencies. In the lab, the distance is a few feet and so the frequencies are much higher. Just what type of waves we are seeing in these laboratory data is a bit unclear to me, especially since I don't know the geometry - probably mostly surface waves. \nAlso, I don't think we are looking here at destructive waves - note that the seismic amplitude at the failure is small - no spikes there. The paper calls them \"slow earthquakes\".",
      "votes": null
    },
    {
      "id": "456573",
      "postDate": "01/16/2019 04:33:43",
      "content": "<p>I tried to looked at the sample spectrum below 1k Hz, but failed... too blur</p>",
      "rawMarkdown": "I tried to looked at the sample spectrum below 1k Hz, but failed... too blur",
      "votes": null
    },
    {
      "id": "456575",
      "postDate": "01/16/2019 04:36:48",
      "content": "<p>if, say, even in lab condition, we are expecting small frequency ones, then, aggregated statistics/features might end up being more informative than the features from windows within windows...</p>",
      "rawMarkdown": "if, say, even in lab condition, we are expecting small frequency ones, then, aggregated statistics/features might end up being more informative than the features from windows within windows...",
      "votes": null
    },
    {
      "id": "456715",
      "postDate": "01/16/2019 11:15:38",
      "content": "<p>My thoughts are here. I am more and more convinced there's a mistake in competition description.\n<a href=\"https://www.kaggle.com/c/LANL-Earthquake-Prediction/discussion/77708\">https://www.kaggle.com/c/LANL-Earthquake-Prediction/discussion/77708</a></p>",
      "rawMarkdown": "My thoughts are here. I am more and more convinced there's a mistake in competition description.\nhttps://www.kaggle.com/c/LANL-Earthquake-Prediction/discussion/77708",
      "votes": null
    },
    {
      "id": "457071",
      "postDate": "01/16/2019 22:41:56",
      "content": "<p>umm i'm not sure about those. \nbut considering such experiment/simulation is pretty standard in earth science, i choose to trust the description. it is worth mentioning though, i didn't think of that before.</p>",
      "rawMarkdown": "umm i'm not sure about those. \nbut considering such experiment/simulation is pretty standard in earth science, i choose to trust the description. it is worth mentioning though, i didn't think of that before.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 456570,
      "author_name": "petewills",
      "author_url": "",
      "post_date": "01/16/2019 04:21:43",
      "content": "<p>Real world earthquake monitoring uses much lower sample rates - maybe 20 Hz max. The waves travel hundreds of miles and are subject to plenty of attenuation, especially the high frequencies. In the lab, the distance is a few feet and so the frequencies are much higher. Just what type of waves we are seeing in these laboratory data is a bit unclear to me, especially since I don't know the geometry - probably mostly surface waves. \nAlso, I don't think we are looking here at destructive waves - note that the seismic amplitude at the failure is small - no spikes there. The paper calls them \"slow earthquakes\".</p>",
      "votes": null,
      "replies": [
        {
          "id": 456573,
          "author_name": "tclf90",
          "author_url": "",
          "post_date": "01/16/2019 04:33:43",
          "content": "<p>I tried to looked at the sample spectrum below 1k Hz, but failed... too blur</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 456575,
          "author_name": "tclf90",
          "author_url": "",
          "post_date": "01/16/2019 04:36:48",
          "content": "<p>if, say, even in lab condition, we are expecting small frequency ones, then, aggregated statistics/features might end up being more informative than the features from windows within windows...</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 456715,
      "author_name": "davids1992",
      "author_url": "",
      "post_date": "01/16/2019 11:15:38",
      "content": "<p>My thoughts are here. I am more and more convinced there's a mistake in competition description.\n<a href=\"https://www.kaggle.com/c/LANL-Earthquake-Prediction/discussion/77708\">https://www.kaggle.com/c/LANL-Earthquake-Prediction/discussion/77708</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 457071,
          "author_name": "tclf90",
          "author_url": "",
          "post_date": "01/16/2019 22:41:56",
          "content": "<p>umm i'm not sure about those. \nbut considering such experiment/simulation is pretty standard in earth science, i choose to trust the description. it is worth mentioning though, i didn't think of that before.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "456522": "Hi all,\n\nThe sampling rate is 4 MHz which looks very high to me. \nI don't know if this is a common sampling rate deployed in \nthe real earthquake monitoring task, nor do I know if this is common\nfrequency to 'see' the destructive waves (in lab conditions). \n\nBeing high is not bad. Resolution thing needs to be just right.\n\nSupposedly,  those major destructive waves, indeed, reside in some\nlower frequency bands. What we will need is to zoom-in our spectral \ngraph, especially the lower chunk.\n\nThat is to require a higher resolution in the frequency domain in whichever\nthe transformation we use e.g. STFT\n\nJust realize this. I'm not sure if this will lead to a better modeling strategy.\nIf anyone knows about this, feel free to comment!",
    "456570": "Real world earthquake monitoring uses much lower sample rates - maybe 20 Hz max. The waves travel hundreds of miles and are subject to plenty of attenuation, especially the high frequencies. In the lab, the distance is a few feet and so the frequencies are much higher. Just what type of waves we are seeing in these laboratory data is a bit unclear to me, especially since I don't know the geometry - probably mostly surface waves. \nAlso, I don't think we are looking here at destructive waves - note that the seismic amplitude at the failure is small - no spikes there. The paper calls them \"slow earthquakes\".",
    "456573": "I tried to looked at the sample spectrum below 1k Hz, but failed... too blur",
    "456575": "if, say, even in lab condition, we are expecting small frequency ones, then, aggregated statistics/features might end up being more informative than the features from windows within windows...",
    "456715": "My thoughts are here. I am more and more convinced there's a mistake in competition description.\nhttps://www.kaggle.com/c/LANL-Earthquake-Prediction/discussion/77708",
    "457071": "umm i'm not sure about those. \nbut considering such experiment/simulation is pretty standard in earth science, i choose to trust the description. it is worth mentioning though, i didn't think of that before."
  },
  "source": "meta"
}