{
  "id": 177412,
  "title": "Validation loss doesn't decrease as training loss does",
  "url": "/competitions/landmark-recognition-2020/discussion/177412",
  "author_name": "Jiahui Ma",
  "post_date": "2020-08-25T20:16:22.081000",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I randomly choose 80000 images from the entire training file, and split it for training/validation. The model I use is pre-trained network EfficientNetB7. I encountered the situation where the training loss keeps on dropping, but not the validation loss after a few epochs. I have tried to change the split rate, learning rate…, but it happened everytime. I attached a screenshot for this.</p>\n<p>Anyone has any suggestions? Is it an overfitting? Please advise me!!!!</p>",
  "messages": [
    {
      "id": 985555,
      "postDate": "2020-08-25T20:16:22.080Z",
      "content": "<p>I randomly choose 80000 images from the entire training file, and split it for training/validation. The model I use is pre-trained network EfficientNetB7. I encountered the situation where the training loss keeps on dropping, but not the validation loss after a few epochs. I have tried to change the split rate, learning rate…, but it happened everytime. I attached a screenshot for this.</p>\n<p>Anyone has any suggestions? Is it an overfitting? Please advise me!!!!</p>",
      "rawMarkdown": "I randomly choose 80000 images from the entire training file, and split it for training/validation. The model I use is pre-trained network EfficientNetB7. I encountered the situation where the training loss keeps on dropping, but not the validation loss after a few epochs. I have tried to change the split rate, learning rate..., but it happened everytime. I attached a screenshot for this.\n\nAnyone has any suggestions? Is it an overfitting? Please advise me!!!!",
      "votes": 1
    },
    {
      "id": 986725,
      "postDate": "2020-08-26T18:20:20.003Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 986725,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-08-26T18:20:20.003000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "985555": "I randomly choose 80000 images from the entire training file, and split it for training/validation. The model I use is pre-trained network EfficientNetB7. I encountered the situation where the training loss keeps on dropping, but not the validation loss after a few epochs. I have tried to change the split rate, learning rate..., but it happened everytime. I attached a screenshot for this.\n\nAnyone has any suggestions? Is it an overfitting? Please advise me!!!!",
    "986725": ""
  }
}