{
  "id": 218352,
  "title": "Last week tips, suggestions and advice!",
  "url": "/competitions/rfcx-species-audio-detection/discussion/218352",
  "author_name": "Kamal Das",
  "post_date": "2021-02-10T08:30:09.290000",
  "votes": -7,
  "comment_count": 8,
  "views": 0,
  "content": "<p>Hi all,</p>\n<p>As we enter the last week of the competition, <br>\nany tips, suggestions and advice are welcome and highly appreciated!! 🙏🙌🖖</p>\n<p>Hoping to learn from the fellow Kagglers will suggest what augmentations/ technique or method helped them improve and what is best to try out in the last week!!</p>\n<p>thanks in advance for your suggestions and advice!</p>",
  "messages": [
    {
      "id": 1201647,
      "postDate": "2021-02-15T14:43:39.550Z",
      "content": "<p>if you time, i have a bold suggestion …. but may not work</p>\n<ul>\n<li>why don't try self-supervised learning: use both test and train as training data and then predict future frames of spectrogram with past frames as input. this is representation learning. maybe this can make really good embedding features</li>\n</ul>",
      "rawMarkdown": "if you time, i have a bold suggestion .... but may not work\n- why don't try self-supervised learning: use both test and train as training data and then predict future frames of spectrogram with past frames as input. this is representation learning. maybe this can make really good embedding features",
      "votes": 1,
      "replies": [
        {
          "id": 1201755,
          "postDate": "2021-02-15T16:23:22.177Z",
          "content": "<p>Thanks <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> !! I am not sure how to go about it… any suggestions/reading recommnedations? Will read and learn about it, though I will not have time to apply it to this competition.</p>",
          "rawMarkdown": "Thanks @hengck23 !! I am not sure how to go about it... any suggestions/reading recommnedations? Will read and learn about it, though I will not have time to apply it to this competition."
        },
        {
          "id": 1205031,
          "postDate": "2021-02-16T13:43:45.427Z",
          "content": "<p>e.g. <a href=\"https://openreview.net/forum?id=r1xBk6NKwr\" target=\"_blank\">https://openreview.net/forum?id=r1xBk6NKwr</a><br>\nLearning audio representations with self-supervision</p>",
          "rawMarkdown": "e.g. https://openreview.net/forum?id=r1xBk6NKwr\nLearning audio representations with self-supervision",
          "votes": 1
        },
        {
          "id": 1205033,
          "postDate": "2021-02-16T13:46:53.437Z",
          "content": "<p>Thanks <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> !!</p>",
          "rawMarkdown": "Thanks @hengck23 !!"
        }
      ]
    },
    {
      "id": 1194541,
      "postDate": "2021-02-10T08:30:09.290Z",
      "content": "<p>Hi all,</p>\n<p>As we enter the last week of the competition, <br>\nany tips, suggestions and advice are welcome and highly appreciated!! 🙏🙌🖖</p>\n<p>Hoping to learn from the fellow Kagglers will suggest what augmentations/ technique or method helped them improve and what is best to try out in the last week!!</p>\n<p>thanks in advance for your suggestions and advice!</p>",
      "rawMarkdown": "Hi all,\n\nAs we enter the last week of the competition, \nany tips, suggestions and advice are welcome and highly appreciated!! 🙏🙌🖖\n\nHoping to learn from the fellow Kagglers will suggest what augmentations/ technique or method helped them improve and what is best to try out in the last week!!\n\nthanks in advance for your suggestions and advice!",
      "votes": -8
    },
    {
      "id": 1201157,
      "postDate": "2021-02-15T07:48:16.553Z",
      "rawMarkdown": "",
      "votes": -4,
      "isDeleted": true,
      "replies": [
        {
          "id": 1201461,
          "postDate": "2021-02-15T12:11:03.830Z",
          "content": "<p>Did you implement your bot with <code>np.abs(score)</code> by chance?</p>",
          "rawMarkdown": "Did you implement your bot with `np.abs(score)` by chance?"
        },
        {
          "id": 1201489,
          "postDate": "2021-02-15T12:34:05.413Z",
          "content": "<p>😆+10randomcharacters</p>",
          "rawMarkdown": "😆+10randomcharacters"
        },
        {
          "id": 1205897,
          "postDate": "2021-02-17T03:22:13.430Z",
          "content": "<p>It works                              </p>",
          "rawMarkdown": "It works                              "
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 1201647,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "2021-02-15T14:43:39.550000",
      "content": "<p>if you time, i have a bold suggestion …. but may not work</p>\n<ul>\n<li>why don't try self-supervised learning: use both test and train as training data and then predict future frames of spectrogram with past frames as input. this is representation learning. maybe this can make really good embedding features</li>\n</ul>",
      "votes": 1,
      "replies": [
        {
          "id": 1201755,
          "author_name": "Kamal Das",
          "author_url": "",
          "post_date": "2021-02-15T16:23:22.177000",
          "content": "<p>Thanks <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> !! I am not sure how to go about it… any suggestions/reading recommnedations? Will read and learn about it, though I will not have time to apply it to this competition.</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1205031,
          "author_name": "hengck23",
          "author_url": "",
          "post_date": "2021-02-16T13:43:45.427000",
          "content": "<p>e.g. <a href=\"https://openreview.net/forum?id=r1xBk6NKwr\" target=\"_blank\">https://openreview.net/forum?id=r1xBk6NKwr</a><br>\nLearning audio representations with self-supervision</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 1205033,
          "author_name": "Kamal Das",
          "author_url": "",
          "post_date": "2021-02-16T13:46:53.437000",
          "content": "<p>Thanks <a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> !!</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1201157,
      "author_name": "",
      "author_url": "",
      "post_date": "2021-02-15T07:48:16.553000",
      "content": "",
      "votes": -4,
      "replies": [
        {
          "id": 1201461,
          "author_name": "Psi",
          "author_url": "",
          "post_date": "2021-02-15T12:11:03.830000",
          "content": "<p>Did you implement your bot with <code>np.abs(score)</code> by chance?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1201489,
          "author_name": "Dieter",
          "author_url": "",
          "post_date": "2021-02-15T12:34:05.413000",
          "content": "<p>😆+10randomcharacters</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1205897,
          "author_name": "Mohammed Rizin V K",
          "author_url": "",
          "post_date": "2021-02-17T03:22:13.430000",
          "content": "<p>It works                              </p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1201647": "if you time, i have a bold suggestion .... but may not work\n- why don't try self-supervised learning: use both test and train as training data and then predict future frames of spectrogram with past frames as input. this is representation learning. maybe this can make really good embedding features",
    "1194541": "Hi all,\n\nAs we enter the last week of the competition, \nany tips, suggestions and advice are welcome and highly appreciated!! 🙏🙌🖖\n\nHoping to learn from the fellow Kagglers will suggest what augmentations/ technique or method helped them improve and what is best to try out in the last week!!\n\nthanks in advance for your suggestions and advice!",
    "1201157": ""
  }
}