{
  "id": 441467,
  "title": "Effectiveness of downsampling",
  "url": "/competitions/child-mind-institute-detect-sleep-states/discussion/441467",
  "author_name": "",
  "post_date": "2023-09-18T23:20:41.492205Z",
  "votes": 5,
  "comment_count": 8,
  "views": 0,
  "content": "<p>Measuring once every 5 seconds results in a very large amount of data. What about downsampling the data to 30 seconds or once per minute?<br>\nAsk for opinions.</p>",
  "messages": [
    {
      "id": "2445541",
      "postDate": "09/18/2023 23:20:41",
      "content": "<p>Measuring once every 5 seconds results in a very large amount of data. What about downsampling the data to 30 seconds or once per minute?<br>\nAsk for opinions.</p>",
      "rawMarkdown": "Measuring once every 5 seconds results in a very large amount of data. What about downsampling the data to 30 seconds or once per minute?\nAsk for opinions.",
      "votes": null
    },
    {
      "id": "2445551",
      "postDate": "09/18/2023 23:51:06",
      "content": "<p>Interesting approach, possible to do and evaluate the score though. Yet I find that more data will result in better model, so I think personally I will stick to 5s sampling (very glad if there is shorter sampling like 3s).</p>",
      "rawMarkdown": "Interesting approach, possible to do and evaluate the score though. Yet I find that more data will result in better model, so I think personally I will stick to 5s sampling (very glad if there is shorter sampling like 3s).",
      "votes": null
    },
    {
      "id": "2445566",
      "postDate": "09/19/2023 00:22:29",
      "content": "<p>Thank you for your opinion <a href=\"https://www.kaggle.com/rimbax\" target=\"_blank\">@rimbax</a> <br>\nDo you think it is a good idea to use a moving average like many people do?</p>",
      "rawMarkdown": "Thank you for your opinion @rimbax \nDo you think it is a good idea to use a moving average like many people do?",
      "votes": null
    },
    {
      "id": "2445577",
      "postDate": "09/19/2023 00:47:36",
      "content": "<p>Yes sure, it acts as a smoothing method whether to use it on a feature like anglez/enmo, or even to use it to smooth the prediction score. It is also proven to be good performance in public notebooks.</p>",
      "rawMarkdown": "Yes sure, it acts as a smoothing method whether to use it on a feature like anglez/enmo, or even to use it to smooth the prediction score. It is also proven to be good performance in public notebooks.",
      "votes": null
    },
    {
      "id": "2445582",
      "postDate": "09/19/2023 00:56:48",
      "content": "<p>thank you!<br>\nI will continue my current activities!<br>\nGood luck!</p>",
      "rawMarkdown": "thank you!\nI will continue my current activities!\nGood luck!",
      "votes": null
    },
    {
      "id": "2445589",
      "postDate": "09/19/2023 01:04:51",
      "content": "<p>Yup, good luck for you too! <a href=\"https://www.kaggle.com/ryotapy\" target=\"_blank\">@ryotapy</a> </p>",
      "rawMarkdown": "Yup, good luck for you too! @ryotapy",
      "votes": null
    },
    {
      "id": "2445613",
      "postDate": "09/19/2023 01:28:32",
      "content": "<p>I agree with you and would like to give it a try. As I discussed <a href=\"https://www.kaggle.com/competitions/child-mind-institute-detect-sleep-states/discussion/438438#2439934/discussion/438438#2439934\" target=\"_blank\">here</a>, it may have something to do with evaluation metrics. If successful, it will give an advantage in data handling.</p>",
      "rawMarkdown": "I agree with you and would like to give it a try. As I discussed [here](https://www.kaggle.com/competitions/child-mind-institute-detect-sleep-states/discussion/438438#2439934/discussion/438438#2439934), it may have something to do with evaluation metrics. If successful, it will give an advantage in data handling.",
      "votes": null
    },
    {
      "id": "2445699",
      "postDate": "09/19/2023 03:22:23",
      "content": "<p>I'm trying it out too! good luck <a href=\"https://www.kaggle.com/mitsuyasuhoshino\" target=\"_blank\">@mitsuyasuhoshino</a> </p>",
      "rawMarkdown": "I'm trying it out too! good luck @mitsuyasuhoshino",
      "votes": null
    },
    {
      "id": "2445862",
      "postDate": "09/19/2023 06:04:13",
      "content": "<p>Dear <a href=\"https://www.kaggle.com/ryotapy\" target=\"_blank\">@ryotapy</a> </p>\n<p>As the famous saying goes:</p>\n<blockquote>\n  <p>\"<em>I am convinced that about one-half the money I spend for advertising is wasted, but I have never been able to decide which half.</em>\"</p>\n</blockquote>\n<p>In this competition we are tasked with identifying, more or less, one sleep <code>onset</code> event and one <code>waking</code> event per day with as much precision as possible (say to within a few minutes) so almost all of the other data in the day is not needed at all… …but on the other hand, having finer grained the data around the specific events is highly beneficial…</p>\n<p>All the best,<br>\ncarl</p>",
      "rawMarkdown": "Dear @ryotapy \n\nAs the famous saying goes:\n\n> \"*I am convinced that about one-half the money I spend for advertising is wasted, but I have never been able to decide which half.*\"\n\nIn this competition we are tasked with identifying, more or less, one sleep `onset` event and one `waking` event per day with as much precision as possible (say to within a few minutes) so almost all of the other data in the day is not needed at all... ...but on the other hand, having finer grained the data around the specific events is highly beneficial...\n\nAll the best,\ncarl",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2445551,
      "author_name": "rimbax",
      "author_url": "",
      "post_date": "09/18/2023 23:51:06",
      "content": "<p>Interesting approach, possible to do and evaluate the score though. Yet I find that more data will result in better model, so I think personally I will stick to 5s sampling (very glad if there is shorter sampling like 3s).</p>",
      "votes": null,
      "replies": [
        {
          "id": 2445566,
          "author_name": "ryotapy",
          "author_url": "",
          "post_date": "09/19/2023 00:22:29",
          "content": "<p>Thank you for your opinion <a href=\"https://www.kaggle.com/rimbax\" target=\"_blank\">@rimbax</a> <br>\nDo you think it is a good idea to use a moving average like many people do?</p>",
          "votes": null,
          "replies": [
            {
              "id": 2445577,
              "author_name": "rimbax",
              "author_url": "",
              "post_date": "09/19/2023 00:47:36",
              "content": "<p>Yes sure, it acts as a smoothing method whether to use it on a feature like anglez/enmo, or even to use it to smooth the prediction score. It is also proven to be good performance in public notebooks.</p>",
              "votes": null,
              "replies": [
                {
                  "id": 2445582,
                  "author_name": "ryotapy",
                  "author_url": "",
                  "post_date": "09/19/2023 00:56:48",
                  "content": "<p>thank you!<br>\nI will continue my current activities!<br>\nGood luck!</p>",
                  "votes": null,
                  "replies": [
                    {
                      "id": 2445589,
                      "author_name": "rimbax",
                      "author_url": "",
                      "post_date": "09/19/2023 01:04:51",
                      "content": "<p>Yup, good luck for you too! <a href=\"https://www.kaggle.com/ryotapy\" target=\"_blank\">@ryotapy</a> </p>",
                      "votes": null,
                      "replies": []
                    }
                  ]
                }
              ]
            }
          ]
        }
      ]
    },
    {
      "id": 2445613,
      "author_name": "mitsuyasuhoshino",
      "author_url": "",
      "post_date": "09/19/2023 01:28:32",
      "content": "<p>I agree with you and would like to give it a try. As I discussed <a href=\"https://www.kaggle.com/competitions/child-mind-institute-detect-sleep-states/discussion/438438#2439934/discussion/438438#2439934\" target=\"_blank\">here</a>, it may have something to do with evaluation metrics. If successful, it will give an advantage in data handling.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2445699,
          "author_name": "ryotapy",
          "author_url": "",
          "post_date": "09/19/2023 03:22:23",
          "content": "<p>I'm trying it out too! good luck <a href=\"https://www.kaggle.com/mitsuyasuhoshino\" target=\"_blank\">@mitsuyasuhoshino</a> </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2445862,
      "author_name": "carlmcbrideellis",
      "author_url": "",
      "post_date": "09/19/2023 06:04:13",
      "content": "<p>Dear <a href=\"https://www.kaggle.com/ryotapy\" target=\"_blank\">@ryotapy</a> </p>\n<p>As the famous saying goes:</p>\n<blockquote>\n  <p>\"<em>I am convinced that about one-half the money I spend for advertising is wasted, but I have never been able to decide which half.</em>\"</p>\n</blockquote>\n<p>In this competition we are tasked with identifying, more or less, one sleep <code>onset</code> event and one <code>waking</code> event per day with as much precision as possible (say to within a few minutes) so almost all of the other data in the day is not needed at all… …but on the other hand, having finer grained the data around the specific events is highly beneficial…</p>\n<p>All the best,<br>\ncarl</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2445541": "Measuring once every 5 seconds results in a very large amount of data. What about downsampling the data to 30 seconds or once per minute?\nAsk for opinions.",
    "2445551": "Interesting approach, possible to do and evaluate the score though. Yet I find that more data will result in better model, so I think personally I will stick to 5s sampling (very glad if there is shorter sampling like 3s).",
    "2445566": "Thank you for your opinion @rimbax \nDo you think it is a good idea to use a moving average like many people do?",
    "2445577": "Yes sure, it acts as a smoothing method whether to use it on a feature like anglez/enmo, or even to use it to smooth the prediction score. It is also proven to be good performance in public notebooks.",
    "2445582": "thank you!\nI will continue my current activities!\nGood luck!",
    "2445589": "Yup, good luck for you too! @ryotapy",
    "2445613": "I agree with you and would like to give it a try. As I discussed [here](https://www.kaggle.com/competitions/child-mind-institute-detect-sleep-states/discussion/438438#2439934/discussion/438438#2439934), it may have something to do with evaluation metrics. If successful, it will give an advantage in data handling.",
    "2445699": "I'm trying it out too! good luck @mitsuyasuhoshino",
    "2445862": "Dear @ryotapy \n\nAs the famous saying goes:\n\n> \"*I am convinced that about one-half the money I spend for advertising is wasted, but I have never been able to decide which half.*\"\n\nIn this competition we are tasked with identifying, more or less, one sleep `onset` event and one `waking` event per day with as much precision as possible (say to within a few minutes) so almost all of the other data in the day is not needed at all... ...but on the other hand, having finer grained the data around the specific events is highly beneficial...\n\nAll the best,\ncarl"
  },
  "source": "meta"
}