{
  "id": 410225,
  "title": "Could this be design if the model",
  "url": "/competitions/asl-fingerspelling/discussion/410225",
  "author_name": "dewa sahu",
  "post_date": "2023-05-14T12:33:25.748000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I'm thinking about architecturing the model to get least loss and more accurate prediction</p>\n<ul>\n<li>model dynamics: as i would have to make the model inputs to be dynamic as some of the features are <code>Nan</code> and every sequence would be having different number of features row</li>\n<li>or i can add the label to the end of row which will make the training more robust</li>\n</ul>\n<p>Im attaching a image as basic architecture of the model could someone suggest <a href=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F11555464%2Fd80b0679ba52458a06a8bc3bc1b2fc22%2Fnull.png?generation=1684067079382562&amp;alt=media\" target=\"_blank\">Image</a></p>",
  "messages": [
    {
      "id": 2258712,
      "postDate": "2023-05-14T12:33:25.750Z",
      "content": "<p>I'm thinking about architecturing the model to get least loss and more accurate prediction</p>\n<ul>\n<li>model dynamics: as i would have to make the model inputs to be dynamic as some of the features are <code>Nan</code> and every sequence would be having different number of features row</li>\n<li>or i can add the label to the end of row which will make the training more robust</li>\n</ul>\n<p>Im attaching a image as basic architecture of the model could someone suggest <a href=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F11555464%2Fd80b0679ba52458a06a8bc3bc1b2fc22%2Fnull.png?generation=1684067079382562&amp;alt=media\" target=\"_blank\">Image</a></p>",
      "rawMarkdown": "I'm thinking about architecturing the model to get least loss and more accurate prediction\n- model dynamics: as i would have to make the model inputs to be dynamic as some of the features are `Nan` and every sequence would be having different number of features row\n- or i can add the label to the end of row which will make the training more robust\n\nIm attaching a image as basic architecture of the model could someone suggest [Image](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F11555464%2Fd80b0679ba52458a06a8bc3bc1b2fc22%2Fnull.png?generation=1684067079382562&alt=media)\n"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2258712": "I'm thinking about architecturing the model to get least loss and more accurate prediction\n- model dynamics: as i would have to make the model inputs to be dynamic as some of the features are `Nan` and every sequence would be having different number of features row\n- or i can add the label to the end of row which will make the training more robust\n\nIm attaching a image as basic architecture of the model could someone suggest [Image](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F11555464%2Fd80b0679ba52458a06a8bc3bc1b2fc22%2Fnull.png?generation=1684067079382562&alt=media)\n"
  }
}