{
  "id": 168399,
  "title": "how to map my train model features to cover the test model feature?",
  "url": "/competitions/prostate-cancer-grade-assessment/discussion/168399",
  "author_name": "",
  "post_date": "2020-07-20T13:44:14.525000",
  "votes": -7,
  "comment_count": 0,
  "views": 0,
  "content": "<p>i find my model can train to a high score , such as 0.92,0.93.\nbut the commit score maybe  0.1.\ni  adjust my train policy,  but time has spent.\ni think just the dataset has little feature vector in the test dataset.\nso i should construct more different dataset from train images.\ncommit score:0.362\ntrain_score:0.87</p>",
  "messages": [
    {
      "id": 936739,
      "postDate": "2020-07-20T13:44:14.527Z",
      "content": "<p>i find my model can train to a high score , such as 0.92,0.93.\nbut the commit score maybe  0.1.\ni  adjust my train policy,  but time has spent.\ni think just the dataset has little feature vector in the test dataset.\nso i should construct more different dataset from train images.\ncommit score:0.362\ntrain_score:0.87</p>",
      "rawMarkdown": "i find my model can train to a high score , such as 0.92,0.93.\nbut the commit score maybe  0.1.\ni  adjust my train policy,  but time has spent.\ni think just the dataset has little feature vector in the test dataset.\nso i should construct more different dataset from train images.\ncommit score:0.362\ntrain_score:0.87",
      "votes": -7
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "936739": "i find my model can train to a high score , such as 0.92,0.93.\nbut the commit score maybe  0.1.\ni  adjust my train policy,  but time has spent.\ni think just the dataset has little feature vector in the test dataset.\nso i should construct more different dataset from train images.\ncommit score:0.362\ntrain_score:0.87"
  }
}