{
  "id": 584008,
  "title": "Some statistics related info of the OpenFWI dataset.",
  "url": "/competitions/waveform-inversion/discussion/584008",
  "author_name": "",
  "post_date": "2025-06-11T05:35:44.626000",
  "votes": 4,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Helped me while experimenting normalization techniques, interpolation and many more.  I didn't experiment much with the actual models itself but normalizing all families with an overall cumulative std, variance etc seems to help. And while inference make sure to use the same values for de-normalizing. Could help in making a unified model for all the families.  </p>\n<pre><code>{\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    }\n}\n</code></pre>",
  "messages": [
    {
      "id": 3221512,
      "postDate": "2025-06-11T05:35:44.627Z",
      "content": "<p>Helped me while experimenting normalization techniques, interpolation and many more.  I didn't experiment much with the actual models itself but normalizing all families with an overall cumulative std, variance etc seems to help. And while inference make sure to use the same values for de-normalizing. Could help in making a unified model for all the families.  </p>\n<pre><code>{\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    },\n     {\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         ,\n         \n    }\n}\n</code></pre>",
      "rawMarkdown": "Helped me while experimenting normalization techniques, interpolation and many more.  I didn't experiment much with the actual models itself but normalizing all families with an overall cumulative std, variance etc seems to help. And while inference make sure to use the same values for de-normalizing. Could help in making a unified model for all the families.  \n```\n{\n    \"flatvel-a\": {\n        \"data_min\": -26.95,\n        \"data_max\": 52.77,\n        \"data_mean\": -3.3804e-05,\n        \"data_std\": 1.4797,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 2782.0442,\n        \"label_std\": 786.1557,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"curvevel-a\": {\n        \"data_min\": -27.11,\n        \"data_max\": 55.10,\n        \"data_mean\": -5.0961e-05,\n        \"data_std\": 1.4870,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"file_size\": 500,\n        \"label_mean\": 2788.1562,\n        \"label_std\": 794.2432,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"flatvel-b\": {\n        \"data_min\": -27.17,\n        \"data_max\": 56.05,\n        \"data_mean\": -0.0002,\n        \"data_std\": 1.7832,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 3001.3389,\n        \"label_std\": 866.6636,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"curvevel-b\": {\n        \"data_min\": -29.04,\n        \"data_max\": 57.03,\n        \"data_mean\": -0.0002,\n        \"data_std\": 1.7648,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 3000.5669,\n        \"label_std\": 865.4404,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n\t\"flatfault-a\": {\n        \"data_min\": -26.10,\n        \"data_max\": 50.86,\n        \"data_mean\": -0.00043503073,\n        \"data_std\": 1.5410482,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 3088.6873,\n        \"label_std\": 855.37024,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"curvefault-a\": {\n        \"data_min\": -26.48,\n        \"data_max\": 52.32,\n        \"data_mean\": -0.00045603843,\n        \"data_std\": 1.5448948,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 3082.6616,\n        \"label_std\": 852.38995,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"flatfault-b\": {\n        \"data_min\": -24.86,\n        \"data_max\": 50.28,\n        \"data_mean\": -0.0001,\n        \"data_std\": 1.4952,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"file_size\": 500,\n        \"label_mean\": 3055.4231,\n        \"label_std\": 875.8992,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"curvefault-b\": {\n        \"data_min\": -24.93,\n        \"data_max\": 50.98,\n        \"data_mean\": -8.882544e-05,\n        \"data_std\": 1.50228,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 3035.5576,\n        \"label_std\": 890.48785,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"style-a\": {\n        \"data_min\": -24.96,\n        \"data_max\": 48.93,\n        \"data_mean\": 0.00024991733,\n        \"data_std\": 1.4618,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 2728.5144,\n        \"label_std\": 665.83215,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"style-b\": {\n        \"data_min\": -23.76,\n        \"data_max\": 46.01,\n        \"data_mean\": 0.00013498258,\n        \"data_std\": 1.4579,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 2837.3164,\n        \"label_std\": 637.6763,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"flatvel-tutorial\": {\n        \"data_min\": -26.95,\n        \"data_max\": 52.77,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"file_size\": 120,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    }\n}\n```",
      "votes": 4
    },
    {
      "id": 3225186,
      "postDate": "2025-06-16T04:39:44.657Z",
      "content": "<p>Thanks for sharing your insights! Maybe that’s a great point about normalizing all families using overall cumulative statistics.</p>",
      "rawMarkdown": "Thanks for sharing your insights! Maybe that’s a great point about normalizing all families using overall cumulative statistics.",
      "votes": 1,
      "replies": [
        {
          "id": 3225211,
          "postDate": "2025-06-16T05:40:42.683Z",
          "rawMarkdown": "",
          "isDeleted": true
        }
      ]
    },
    {
      "id": 3225184,
      "postDate": "2025-06-16T04:39:00.833Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 3225186,
      "author_name": "Adam",
      "author_url": "",
      "post_date": "2025-06-16T04:39:44.657000",
      "content": "<p>Thanks for sharing your insights! Maybe that’s a great point about normalizing all families using overall cumulative statistics.</p>",
      "votes": 1,
      "replies": [
        {
          "id": 3225211,
          "author_name": "",
          "author_url": "",
          "post_date": "2025-06-16T05:40:42.683000",
          "content": "",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 3225184,
      "author_name": "",
      "author_url": "",
      "post_date": "2025-06-16T04:39:00.833000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3221512": "Helped me while experimenting normalization techniques, interpolation and many more.  I didn't experiment much with the actual models itself but normalizing all families with an overall cumulative std, variance etc seems to help. And while inference make sure to use the same values for de-normalizing. Could help in making a unified model for all the families.  \n```\n{\n    \"flatvel-a\": {\n        \"data_min\": -26.95,\n        \"data_max\": 52.77,\n        \"data_mean\": -3.3804e-05,\n        \"data_std\": 1.4797,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 2782.0442,\n        \"label_std\": 786.1557,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"curvevel-a\": {\n        \"data_min\": -27.11,\n        \"data_max\": 55.10,\n        \"data_mean\": -5.0961e-05,\n        \"data_std\": 1.4870,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"file_size\": 500,\n        \"label_mean\": 2788.1562,\n        \"label_std\": 794.2432,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"flatvel-b\": {\n        \"data_min\": -27.17,\n        \"data_max\": 56.05,\n        \"data_mean\": -0.0002,\n        \"data_std\": 1.7832,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 3001.3389,\n        \"label_std\": 866.6636,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"curvevel-b\": {\n        \"data_min\": -29.04,\n        \"data_max\": 57.03,\n        \"data_mean\": -0.0002,\n        \"data_std\": 1.7648,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 3000.5669,\n        \"label_std\": 865.4404,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n\t\"flatfault-a\": {\n        \"data_min\": -26.10,\n        \"data_max\": 50.86,\n        \"data_mean\": -0.00043503073,\n        \"data_std\": 1.5410482,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 3088.6873,\n        \"label_std\": 855.37024,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"curvefault-a\": {\n        \"data_min\": -26.48,\n        \"data_max\": 52.32,\n        \"data_mean\": -0.00045603843,\n        \"data_std\": 1.5448948,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 3082.6616,\n        \"label_std\": 852.38995,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"flatfault-b\": {\n        \"data_min\": -24.86,\n        \"data_max\": 50.28,\n        \"data_mean\": -0.0001,\n        \"data_std\": 1.4952,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"file_size\": 500,\n        \"label_mean\": 3055.4231,\n        \"label_std\": 875.8992,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"curvefault-b\": {\n        \"data_min\": -24.93,\n        \"data_max\": 50.98,\n        \"data_mean\": -8.882544e-05,\n        \"data_std\": 1.50228,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 3035.5576,\n        \"label_std\": 890.48785,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"style-a\": {\n        \"data_min\": -24.96,\n        \"data_max\": 48.93,\n        \"data_mean\": 0.00024991733,\n        \"data_std\": 1.4618,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 2728.5144,\n        \"label_std\": 665.83215,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"style-b\": {\n        \"data_min\": -23.76,\n        \"data_max\": 46.01,\n        \"data_mean\": 0.00013498258,\n        \"data_std\": 1.4579,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"label_mean\": 2837.3164,\n        \"label_std\": 637.6763,\n        \"file_size\": 500,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    },\n    \"flatvel-tutorial\": {\n        \"data_min\": -26.95,\n        \"data_max\": 52.77,\n        \"label_min\": 1500,\n        \"label_max\": 4500,\n        \"file_size\": 120,\n        \"nbc\": 120,\n        \"dx\": 10,\n        \"nt\": 1000,\n        \"dt\": 1e-3,\n        \"f\": 15,\n        \"n_grid\": 70,\n        \"ns\": 5,\n        \"ng\": 70,\n        \"sz\": 10,\n        \"gz\": 10\n    }\n}\n```",
    "3225186": "Thanks for sharing your insights! Maybe that’s a great point about normalizing all families using overall cumulative statistics.",
    "3225184": ""
  }
}