{
  "id": 275368,
  "title": "A little trick from a newbie ~ 0.0030 increase in both public and private",
  "url": "/competitions/g2net-gravitational-wave-detection/discussion/275368",
  "author_name": "",
  "post_date": "2021-09-30T05:55:34.080370500Z",
  "votes": 9,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I've added the real and imaginary parts of the CQT to the input, just like that:</p>\n<p><code>CQT = tf.concat([CQT_real[0], CQT_imag[0], CQT[0]], axis=1)</code></p>\n<p>It gave me a substantial boost in both Private and public LB for EffNetB5, which for me is place 65 vs 265.</p>\n<p>I haven't seen that anywhere else.</p>\n<p>Anyway, it has been an amazing way to learn new things, thank you kaggle community!</p>",
  "messages": [
    {
      "id": "1529058",
      "postDate": "09/30/2021 05:55:34",
      "content": "<p>I've added the real and imaginary parts of the CQT to the input, just like that:</p>\n<p><code>CQT = tf.concat([CQT_real[0], CQT_imag[0], CQT[0]], axis=1)</code></p>\n<p>It gave me a substantial boost in both Private and public LB for EffNetB5, which for me is place 65 vs 265.</p>\n<p>I haven't seen that anywhere else.</p>\n<p>Anyway, it has been an amazing way to learn new things, thank you kaggle community!</p>",
      "rawMarkdown": "I've added the real and imaginary parts of the CQT to the input, just like that:\n\n`CQT = tf.concat([CQT_real[0], CQT_imag[0], CQT[0]], axis=1)`\n\nIt gave me a substantial boost in both Private and public LB for EffNetB5, which for me is place 65 vs 265.\n\nI haven't seen that anywhere else.\n\nAnyway, it has been an amazing way to learn new things, thank you kaggle community!",
      "votes": null
    },
    {
      "id": "1529195",
      "postDate": "09/30/2021 08:10:09",
      "content": "<p>Is it better than using magnitude directly? I doubt it frankly.</p>",
      "rawMarkdown": "Is it better than using magnitude directly? I doubt it frankly.",
      "votes": null
    },
    {
      "id": "1529280",
      "postDate": "09/30/2021 09:07:45",
      "content": "<p>0.8738 vs 8768 on EFFNetB5 - private, 8747 vs 8755 - resnet18 public, (8705 vs 8714 oof for the same arch) - this approach always better<br>\nPlease notice that the magnitude is also used.</p>",
      "rawMarkdown": "0.8738 vs 8768 on EFFNetB5 - private, 8747 vs 8755 - resnet18 public, (8705 vs 8714 oof for the same arch) - this approach always better\nPlease notice that the magnitude is also used.",
      "votes": null
    },
    {
      "id": "1529424",
      "postDate": "09/30/2021 11:37:55",
      "content": "<p>In my opinion, 2 components provide more information than just magnitude, like, a vector with a fixed length (magnitude) can rotate everywhere around the origin. But, I could be wrong, please correct me if I was.</p>",
      "rawMarkdown": "In my opinion, 2 components provide more information than just magnitude, like, a vector with a fixed length (magnitude) can rotate everywhere around the origin. But, I could be wrong, please correct me if I was.",
      "votes": null
    },
    {
      "id": "1529425",
      "postDate": "09/30/2021 11:38:24",
      "content": "<p>Sorry, misread you.  You added real and imaginary to the magnitude.  </p>\n<p>I thought you compared both components to only real.  I read too fast.</p>",
      "rawMarkdown": "Sorry, misread you.  You added real and imaginary to the magnitude.  \n\nI thought you compared both components to only real.  I read too fast.",
      "votes": null
    },
    {
      "id": "1559883",
      "postDate": "10/27/2021 08:05:39",
      "content": "<p>Hey All,</p>\n<p>Thank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey <a href=\"https://forms.gle/QP9L16niPexozyhu5\" target=\"_blank\">https://forms.gle/QP9L16niPexozyhu5</a>.</p>\n<p>Thank you all,</p>\n<p>Regards,<br>\nChris</p>",
      "rawMarkdown": "Hey All,\n\nThank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.\n\nThank you all,\n\nRegards,\nChris",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1529195,
      "author_name": "cpmpml",
      "author_url": "",
      "post_date": "09/30/2021 08:10:09",
      "content": "<p>Is it better than using magnitude directly? I doubt it frankly.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1529280,
          "author_name": "noulan",
          "author_url": "",
          "post_date": "09/30/2021 09:07:45",
          "content": "<p>0.8738 vs 8768 on EFFNetB5 - private, 8747 vs 8755 - resnet18 public, (8705 vs 8714 oof for the same arch) - this approach always better<br>\nPlease notice that the magnitude is also used.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1529424,
          "author_name": "shinomoriaoshi",
          "author_url": "",
          "post_date": "09/30/2021 11:37:55",
          "content": "<p>In my opinion, 2 components provide more information than just magnitude, like, a vector with a fixed length (magnitude) can rotate everywhere around the origin. But, I could be wrong, please correct me if I was.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1529425,
          "author_name": "cpmpml",
          "author_url": "",
          "post_date": "09/30/2021 11:38:24",
          "content": "<p>Sorry, misread you.  You added real and imaginary to the magnitude.  </p>\n<p>I thought you compared both components to only real.  I read too fast.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1559883,
      "author_name": "zerafachris",
      "author_url": "",
      "post_date": "10/27/2021 08:05:39",
      "content": "<p>Hey All,</p>\n<p>Thank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey <a href=\"https://forms.gle/QP9L16niPexozyhu5\" target=\"_blank\">https://forms.gle/QP9L16niPexozyhu5</a>.</p>\n<p>Thank you all,</p>\n<p>Regards,<br>\nChris</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1529058": "I've added the real and imaginary parts of the CQT to the input, just like that:\n\n`CQT = tf.concat([CQT_real[0], CQT_imag[0], CQT[0]], axis=1)`\n\nIt gave me a substantial boost in both Private and public LB for EffNetB5, which for me is place 65 vs 265.\n\nI haven't seen that anywhere else.\n\nAnyway, it has been an amazing way to learn new things, thank you kaggle community!",
    "1529195": "Is it better than using magnitude directly? I doubt it frankly.",
    "1529280": "0.8738 vs 8768 on EFFNetB5 - private, 8747 vs 8755 - resnet18 public, (8705 vs 8714 oof for the same arch) - this approach always better\nPlease notice that the magnitude is also used.",
    "1529424": "In my opinion, 2 components provide more information than just magnitude, like, a vector with a fixed length (magnitude) can rotate everywhere around the origin. But, I could be wrong, please correct me if I was.",
    "1529425": "Sorry, misread you.  You added real and imaginary to the magnitude.  \n\nI thought you compared both components to only real.  I read too fast.",
    "1559883": "Hey All,\n\nThank you all for taking part in our competition. The participation has been overwhelmingly positive. We are currently conducting a survey to gauge the demographic and outreach achieved. Kindly spare 2min and fill in this survey https://forms.gle/QP9L16niPexozyhu5.\n\nThank you all,\n\nRegards,\nChris"
  },
  "source": "meta"
}