{
  "id": 75762,
  "title": "beyond oversampling ... meta-learning, k-shot learning and the tail",
  "url": "/competitions/human-protein-atlas-image-classification/discussion/75762",
  "author_name": "",
  "post_date": "2018-12-26T08:50:08.608079600Z",
  "votes": 15,
  "comment_count": 1,
  "views": 0,
  "content": "<p>The objective of this post is to introduce some \"unconventional methods\". </p>\n\n<p>Class imbalance and low train sample count is a key problem to this challenge. \"K-shot learning\" is a solution to this issue and has been researched over the past years. Meta-learning is a method to solved this in the paper discussed below:</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/445349/10932/Slide1.png\" alt=\"enter image description here\"></p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/445349/10933/Slide2.png\" alt=\"enter image description here\"></p>",
  "messages": [
    {
      "id": "445349",
      "postDate": "12/26/2018 08:50:08",
      "content": "<p>The objective of this post is to introduce some \"unconventional methods\". </p>\n\n<p>Class imbalance and low train sample count is a key problem to this challenge. \"K-shot learning\" is a solution to this issue and has been researched over the past years. Meta-learning is a method to solved this in the paper discussed below:</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/445349/10932/Slide1.png\" alt=\"enter image description here\"></p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/445349/10933/Slide2.png\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "The objective of this post is to introduce some \"unconventional methods\". \n\nClass imbalance and low train sample count is a key problem to this challenge. \"K-shot learning\" is a solution to this issue and has been researched over the past years. Meta-learning is a method to solved this in the paper discussed below:\n\n  ![enter image description here][1]\n\n  ![enter image description here][2]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/445349/10932/Slide1.png\n  [2]: https://storage.googleapis.com/kaggle-forum-message-attachments/445349/10933/Slide2.png",
      "votes": null
    },
    {
      "id": "445901",
      "postDate": "12/27/2018 07:27:43",
      "content": "<p>great idea. thanks for sharing.</p>",
      "rawMarkdown": "great idea. thanks for sharing.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 445901,
      "author_name": "dragon229",
      "author_url": "",
      "post_date": "12/27/2018 07:27:43",
      "content": "<p>great idea. thanks for sharing.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "445349": "The objective of this post is to introduce some \"unconventional methods\". \n\nClass imbalance and low train sample count is a key problem to this challenge. \"K-shot learning\" is a solution to this issue and has been researched over the past years. Meta-learning is a method to solved this in the paper discussed below:\n\n  ![enter image description here][1]\n\n  ![enter image description here][2]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/445349/10932/Slide1.png\n  [2]: https://storage.googleapis.com/kaggle-forum-message-attachments/445349/10933/Slide2.png",
    "445901": "great idea. thanks for sharing."
  },
  "source": "meta"
}