{
  "id": 90163,
  "title": "What if we denoise the data?",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/90163",
  "author_name": "",
  "post_date": "2019-04-21T10:05:45.506271200Z",
  "votes": 7,
  "comment_count": 2,
  "views": 0,
  "content": "<p><a href=\"https://www.kaggle.com/zikazika/what-if-we-denoise-the-data\">Denoising</a>\nWhat if we denoise the data as in <a href=\"https://www.kaggle.com/jackvial/dwt-signal-denoising\">DWT</a>. Results are comparable (couple of points of difference).\n2.06 without DWT\n2.1 somethin with DWT</p>\n\n<p>in the worst case we can extract the features that we find relevant/interesting while denoising in hopes that they will be picked up in private LB.</p>",
  "messages": [
    {
      "id": "520566",
      "postDate": "04/21/2019 10:05:45",
      "content": "<p><a href=\"https://www.kaggle.com/zikazika/what-if-we-denoise-the-data\">Denoising</a>\nWhat if we denoise the data as in <a href=\"https://www.kaggle.com/jackvial/dwt-signal-denoising\">DWT</a>. Results are comparable (couple of points of difference).\n2.06 without DWT\n2.1 somethin with DWT</p>\n\n<p>in the worst case we can extract the features that we find relevant/interesting while denoising in hopes that they will be picked up in private LB.</p>",
      "rawMarkdown": "[Denoising](https://www.kaggle.com/zikazika/what-if-we-denoise-the-data)\nWhat if we denoise the data as in [DWT](https://www.kaggle.com/jackvial/dwt-signal-denoising). Results are comparable (couple of points of difference).\n2.06 without DWT\n2.1 somethin with DWT\n\n\n\n\nin the worst case we can extract the features that we find relevant/interesting while denoising in hopes that they will be picked up in private LB.",
      "votes": null
    },
    {
      "id": "522395",
      "postDate": "04/24/2019 11:36:16",
      "content": "<p>I think lowpass-filtering makes sense since the parts &gt;350 kHz seem to be only noise. Might be wrong tho.</p>",
      "rawMarkdown": "I think lowpass-filtering makes sense since the parts &gt;350 kHz seem to be only noise. Might be wrong tho.",
      "votes": null
    },
    {
      "id": "535573",
      "postDate": "05/23/2019 07:05:39",
      "content": "<p>great idea! will try it!\na little bit scared that I may be overfitting a bit :) </p>",
      "rawMarkdown": "great idea! will try it!\na little bit scared that I may be overfitting a bit :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 522395,
      "author_name": "svenhinderer",
      "author_url": "",
      "post_date": "04/24/2019 11:36:16",
      "content": "<p>I think lowpass-filtering makes sense since the parts &gt;350 kHz seem to be only noise. Might be wrong tho.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 535573,
      "author_name": "frtgnn",
      "author_url": "",
      "post_date": "05/23/2019 07:05:39",
      "content": "<p>great idea! will try it!\na little bit scared that I may be overfitting a bit :) </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "520566": "[Denoising](https://www.kaggle.com/zikazika/what-if-we-denoise-the-data)\nWhat if we denoise the data as in [DWT](https://www.kaggle.com/jackvial/dwt-signal-denoising). Results are comparable (couple of points of difference).\n2.06 without DWT\n2.1 somethin with DWT\n\n\n\n\nin the worst case we can extract the features that we find relevant/interesting while denoising in hopes that they will be picked up in private LB.",
    "522395": "I think lowpass-filtering makes sense since the parts &gt;350 kHz seem to be only noise. Might be wrong tho.",
    "535573": "great idea! will try it!\na little bit scared that I may be overfitting a bit :)"
  },
  "source": "meta"
}