{
  "id": 226560,
  "title": "anyone uses LSTM to get labels of patient sequentially?",
  "url": "/competitions/ranzcr-clip-catheter-line-classification/discussion/226560",
  "author_name": "",
  "post_date": "2021-03-17T00:10:15.435597200Z",
  "votes": 8,
  "comment_count": 3,
  "views": 0,
  "content": "<p><img src=\"https://i.ibb.co/g3SfsBd/Selection-344.png\" alt=\"\"></p>\n<p>left, CNN +LSTM : we predict all images for a patient together<br>\nright: CNN results, each image is predicted independently<br>\ngreen: ground truth</p>\n<p>our initial experiment shows it would add diversity and good for ensemble.</p>\n<p>we are thinking of using this method to generate better pseudo label, but unfortunately didn't have time to complete it. </p>\n<p>so i wonder if any team has results to share? i am also interested in the performance of transformer for sequential prediction.</p>",
  "messages": [
    {
      "id": "1241197",
      "postDate": "03/17/2021 00:10:15",
      "content": "<p><img src=\"https://i.ibb.co/g3SfsBd/Selection-344.png\" alt=\"\"></p>\n<p>left, CNN +LSTM : we predict all images for a patient together<br>\nright: CNN results, each image is predicted independently<br>\ngreen: ground truth</p>\n<p>our initial experiment shows it would add diversity and good for ensemble.</p>\n<p>we are thinking of using this method to generate better pseudo label, but unfortunately didn't have time to complete it. </p>\n<p>so i wonder if any team has results to share? i am also interested in the performance of transformer for sequential prediction.</p>",
      "rawMarkdown": "![](https://i.ibb.co/g3SfsBd/Selection-344.png)\n\nleft, CNN +LSTM : we predict all images for a patient together\nright: CNN results, each image is predicted independently\ngreen: ground truth\n\n\nour initial experiment shows it would add diversity and good for ensemble.\n\nwe are thinking of using this method to generate better pseudo label, but unfortunately didn't have time to complete it. \n\nso i wonder if any team has results to share? i am also interested in the performance of transformer for sequential prediction.",
      "votes": null
    },
    {
      "id": "1241200",
      "postDate": "03/17/2021 00:13:19",
      "content": "<p>tried but not work for me so i removed lstm</p>",
      "rawMarkdown": "tried but not work for me so i removed lstm",
      "votes": null
    },
    {
      "id": "1241287",
      "postDate": "03/17/2021 01:31:42",
      "content": "<p>I tried using sequential learning and other meta features and didn't worked for me.</p>",
      "rawMarkdown": "I tried using sequential learning and other meta features and didn't worked for me.",
      "votes": null
    },
    {
      "id": "1243261",
      "postDate": "03/18/2021 05:44:21",
      "content": "<p>I tried LSTM , and transformer as well but it didn't work for me</p>",
      "rawMarkdown": "I tried LSTM , and transformer as well but it didn't work for me",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1241200,
      "author_name": "moewie94",
      "author_url": "",
      "post_date": "03/17/2021 00:13:19",
      "content": "<p>tried but not work for me so i removed lstm</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1241287,
      "author_name": "titericz",
      "author_url": "",
      "post_date": "03/17/2021 01:31:42",
      "content": "<p>I tried using sequential learning and other meta features and didn't worked for me.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1243261,
      "author_name": "tanulsingh077",
      "author_url": "",
      "post_date": "03/18/2021 05:44:21",
      "content": "<p>I tried LSTM , and transformer as well but it didn't work for me</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1241197": "![](https://i.ibb.co/g3SfsBd/Selection-344.png)\n\nleft, CNN +LSTM : we predict all images for a patient together\nright: CNN results, each image is predicted independently\ngreen: ground truth\n\n\nour initial experiment shows it would add diversity and good for ensemble.\n\nwe are thinking of using this method to generate better pseudo label, but unfortunately didn't have time to complete it. \n\nso i wonder if any team has results to share? i am also interested in the performance of transformer for sequential prediction.",
    "1241200": "tried but not work for me so i removed lstm",
    "1241287": "I tried using sequential learning and other meta features and didn't worked for me.",
    "1243261": "I tried LSTM , and transformer as well but it didn't work for me"
  },
  "source": "meta"
}