{
  "id": 55907,
  "title": "Problems with a weighted loss for a multi output Unet",
  "url": "/competitions/data-science-bowl-2018/discussion/55907",
  "author_name": "bilal2vec",
  "post_date": "2018-05-03T02:36:37.148000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi, </p>\n\n<p>I've been trying to train a multioutput unet with two output channels (mask and borders between adjacent nuclei) and a weighted loss like this</p>\n\n<p><code>loss = bceloss(all_preds, all_targets) + dice_loss(mask_preds, mask_targets) + 10 * dice_loss(edges_preds)</code> </p>\n\n<p>The edges outputs from this approach contain all the borders of the nuclei instead of finding the borders between adjacent nuclei. </p>\n\n<p>Could this be happening because the mask and edges losses aren't weighted correctly or could this model be overfitting/underfitting ?</p>\n\n<p>Thanks</p>",
  "messages": [
    {
      "id": 322467,
      "postDate": "2018-05-03T02:36:37.147Z",
      "content": "<p>Hi, </p>\n\n<p>I've been trying to train a multioutput unet with two output channels (mask and borders between adjacent nuclei) and a weighted loss like this</p>\n\n<p><code>loss = bceloss(all_preds, all_targets) + dice_loss(mask_preds, mask_targets) + 10 * dice_loss(edges_preds)</code> </p>\n\n<p>The edges outputs from this approach contain all the borders of the nuclei instead of finding the borders between adjacent nuclei. </p>\n\n<p>Could this be happening because the mask and edges losses aren't weighted correctly or could this model be overfitting/underfitting ?</p>\n\n<p>Thanks</p>",
      "rawMarkdown": "Hi, \n\nI've been trying to train a multioutput unet with two output channels (mask and borders between adjacent nuclei) and a weighted loss like this\n\n`loss = bceloss(all_preds, all_targets) + dice_loss(mask_preds, mask_targets) + 10 * dice_loss(edges_preds)` \n\nThe edges outputs from this approach contain all the borders of the nuclei instead of finding the borders between adjacent nuclei. \n\nCould this be happening because the mask and edges losses aren't weighted correctly or could this model be overfitting/underfitting ?\n\nThanks"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "322467": "Hi, \n\nI've been trying to train a multioutput unet with two output channels (mask and borders between adjacent nuclei) and a weighted loss like this\n\n`loss = bceloss(all_preds, all_targets) + dice_loss(mask_preds, mask_targets) + 10 * dice_loss(edges_preds)` \n\nThe edges outputs from this approach contain all the borders of the nuclei instead of finding the borders between adjacent nuclei. \n\nCould this be happening because the mask and edges losses aren't weighted correctly or could this model be overfitting/underfitting ?\n\nThanks"
  }
}