{
  "id": 227074,
  "title": "Use softmax function in this competition",
  "url": "/competitions/ranzcr-clip-catheter-line-classification/discussion/227074",
  "author_name": "Alien",
  "post_date": "2021-03-18T22:39:47.160000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>My research in this competition is how to mix sigmoid function output and softmax function output. I looked at the last two digits of the results of this competition. There was an improvement of 0.0004~0.001 in the private test set and the public test set. I don't know if this method can improve the results in the 0.971~0.975 range. And whether my idea is correct in this competition. Hope everyone can share your views. And if the competition is multi-class, can I use the sigmoid function to mix model to improve performance?</p>\n<p>I use the softmax function method in this competition.</p>\n<p>The ETT is relatively simple, just add a column ETT000.<br>\nIn the NGT, only a small part of the images will have 2 different labels at the same time, so the image with 2 labels is removed and the NGT000 column is added. Only ensemble the output with the highest ranking for each NGT image.<br>\nFinally, in CVC, I put all combinations with a label, so I will get 8 columns.<br>\nCVC001 CVC010 CVC100 CVC101 CVC011 CVC110 CVC000 CVC111</p>\n<pre><code>submission['CVC - Abnormal'] = sub_df['CVC100'] +sub_df['CVC101'] +sub_df['CVC111'] +sub_df['CVC110']\nsubmission['CVC - Borderline'] = sub_df['CVC110'] + sub_df['CVC010'] + sub_df['CVC111'] + sub_df['CVC011']\nsubmission['CVC - Normal'] = sub_df['CVC101'] +sub_df['CVC001'] +sub_df['CVC111'] +sub_df['CVC011'] \n</code></pre>\n<p>In this way, ETT NGT CVC can be output with softmax function.</p>",
  "messages": [
    {
      "id": 1244316,
      "postDate": "2021-03-18T22:39:47.160Z",
      "content": "<p>My research in this competition is how to mix sigmoid function output and softmax function output. I looked at the last two digits of the results of this competition. There was an improvement of 0.0004~0.001 in the private test set and the public test set. I don't know if this method can improve the results in the 0.971~0.975 range. And whether my idea is correct in this competition. Hope everyone can share your views. And if the competition is multi-class, can I use the sigmoid function to mix model to improve performance?</p>\n<p>I use the softmax function method in this competition.</p>\n<p>The ETT is relatively simple, just add a column ETT000.<br>\nIn the NGT, only a small part of the images will have 2 different labels at the same time, so the image with 2 labels is removed and the NGT000 column is added. Only ensemble the output with the highest ranking for each NGT image.<br>\nFinally, in CVC, I put all combinations with a label, so I will get 8 columns.<br>\nCVC001 CVC010 CVC100 CVC101 CVC011 CVC110 CVC000 CVC111</p>\n<pre><code>submission['CVC - Abnormal'] = sub_df['CVC100'] +sub_df['CVC101'] +sub_df['CVC111'] +sub_df['CVC110']\nsubmission['CVC - Borderline'] = sub_df['CVC110'] + sub_df['CVC010'] + sub_df['CVC111'] + sub_df['CVC011']\nsubmission['CVC - Normal'] = sub_df['CVC101'] +sub_df['CVC001'] +sub_df['CVC111'] +sub_df['CVC011'] \n</code></pre>\n<p>In this way, ETT NGT CVC can be output with softmax function.</p>",
      "rawMarkdown": "My research in this competition is how to mix sigmoid function output and softmax function output. I looked at the last two digits of the results of this competition. There was an improvement of 0.0004~0.001 in the private test set and the public test set. I don't know if this method can improve the results in the 0.971~0.975 range. And whether my idea is correct in this competition. Hope everyone can share your views. And if the competition is multi-class, can I use the sigmoid function to mix model to improve performance?\n\nI use the softmax function method in this competition.\n\nThe ETT is relatively simple, just add a column ETT000.\nIn the NGT, only a small part of the images will have 2 different labels at the same time, so the image with 2 labels is removed and the NGT000 column is added. Only ensemble the output with the highest ranking for each NGT image.\nFinally, in CVC, I put all combinations with a label, so I will get 8 columns.\nCVC001 CVC010 CVC100 CVC101 CVC011 CVC110 CVC000 CVC111\n```\nsubmission['CVC - Abnormal'] = sub_df['CVC100'] +sub_df['CVC101'] +sub_df['CVC111'] +sub_df['CVC110']\nsubmission['CVC - Borderline'] = sub_df['CVC110'] + sub_df['CVC010'] + sub_df['CVC111'] + sub_df['CVC011']\nsubmission['CVC - Normal'] = sub_df['CVC101'] +sub_df['CVC001'] +sub_df['CVC111'] +sub_df['CVC011'] \n```\nIn this way, ETT NGT CVC can be output with softmax function.",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1244316": "My research in this competition is how to mix sigmoid function output and softmax function output. I looked at the last two digits of the results of this competition. There was an improvement of 0.0004~0.001 in the private test set and the public test set. I don't know if this method can improve the results in the 0.971~0.975 range. And whether my idea is correct in this competition. Hope everyone can share your views. And if the competition is multi-class, can I use the sigmoid function to mix model to improve performance?\n\nI use the softmax function method in this competition.\n\nThe ETT is relatively simple, just add a column ETT000.\nIn the NGT, only a small part of the images will have 2 different labels at the same time, so the image with 2 labels is removed and the NGT000 column is added. Only ensemble the output with the highest ranking for each NGT image.\nFinally, in CVC, I put all combinations with a label, so I will get 8 columns.\nCVC001 CVC010 CVC100 CVC101 CVC011 CVC110 CVC000 CVC111\n```\nsubmission['CVC - Abnormal'] = sub_df['CVC100'] +sub_df['CVC101'] +sub_df['CVC111'] +sub_df['CVC110']\nsubmission['CVC - Borderline'] = sub_df['CVC110'] + sub_df['CVC010'] + sub_df['CVC111'] + sub_df['CVC011']\nsubmission['CVC - Normal'] = sub_df['CVC101'] +sub_df['CVC001'] +sub_df['CVC111'] +sub_df['CVC011'] \n```\nIn this way, ETT NGT CVC can be output with softmax function."
  }
}