{
  "id": 207346,
  "title": "StratifiedShuffleSplit reliability",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/207346",
  "author_name": "Fares Lassoued",
  "post_date": "2020-12-29T09:49:23.997000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>As the title says I have limited computing power and want to get the most of a single model. I use StratifiedShuffleSplit from sklearn to get a single stratified train_test_split and my LB and validation scores are very close (single resnet18 model LB:0.875 and Val:0.87336). My question is, should I keep relying on this method given that the data is very noisy or CV is a must to? </p>",
  "messages": [
    {
      "id": 1130764,
      "postDate": "2020-12-29T09:49:23.997Z",
      "content": "<p>As the title says I have limited computing power and want to get the most of a single model. I use StratifiedShuffleSplit from sklearn to get a single stratified train_test_split and my LB and validation scores are very close (single resnet18 model LB:0.875 and Val:0.87336). My question is, should I keep relying on this method given that the data is very noisy or CV is a must to? </p>",
      "rawMarkdown": "As the title says I have limited computing power and want to get the most of a single model. I use StratifiedShuffleSplit from sklearn to get a single stratified train_test_split and my LB and validation scores are very close (single resnet18 model LB:0.875 and Val:0.87336). My question is, should I keep relying on this method given that the data is very noisy or CV is a must to? "
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1130764": "As the title says I have limited computing power and want to get the most of a single model. I use StratifiedShuffleSplit from sklearn to get a single stratified train_test_split and my LB and validation scores are very close (single resnet18 model LB:0.875 and Val:0.87336). My question is, should I keep relying on this method given that the data is very noisy or CV is a must to? "
  }
}