{
  "id": 255209,
  "title": "How`s the effect of Mixup on new data?",
  "url": "/competitions/seti-breakthrough-listen/discussion/255209",
  "author_name": "Hao",
  "post_date": "2021-07-26T11:19:08.219000",
  "votes": 4,
  "comment_count": 2,
  "views": 0,
  "content": "<p>In the history dicussion many people said Mixup is very effective for this competition, like in this <a href=\"https://www.kaggle.com/c/seti-breakthrough-listen/discussion/245152\" target=\"_blank\">discussion</a>(thanks <a href=\"https://www.kaggle.com/ayuraj\" target=\"_blank\">@ayuraj</a>) said ~7% boost in score, but in my experiment on new data I didn`t notice obvious boost with stanard Mixup with 0.5 probability, has any one observe the same? Shall I make my Mixup more aggressive to 1.0 probability?</p>",
  "messages": [
    {
      "id": 1400528,
      "postDate": "2021-07-26T11:19:08.220Z",
      "content": "<p>In the history dicussion many people said Mixup is very effective for this competition, like in this <a href=\"https://www.kaggle.com/c/seti-breakthrough-listen/discussion/245152\" target=\"_blank\">discussion</a>(thanks <a href=\"https://www.kaggle.com/ayuraj\" target=\"_blank\">@ayuraj</a>) said ~7% boost in score, but in my experiment on new data I didn`t notice obvious boost with stanard Mixup with 0.5 probability, has any one observe the same? Shall I make my Mixup more aggressive to 1.0 probability?</p>",
      "rawMarkdown": "In the history dicussion many people said Mixup is very effective for this competition, like in this [discussion](https://www.kaggle.com/c/seti-breakthrough-listen/discussion/245152)(thanks @ayuraj) said ~7% boost in score, but in my experiment on new data I didn`t notice obvious boost with stanard Mixup with 0.5 probability, has any one observe the same? Shall I make my Mixup more aggressive to 1.0 probability?",
      "votes": 4
    },
    {
      "id": 1400947,
      "postDate": "2021-07-26T18:22:51.463Z",
      "content": "<p>I didn't remove mixup maybe I should try.</p>",
      "rawMarkdown": "I didn't remove mixup maybe I should try.",
      "replies": [
        {
          "id": 1401312,
          "postDate": "2021-07-27T07:50:23.380Z",
          "content": "<p>After ajust some sequence of my augmentations, now my Mixup can boost LB+0.01, not good as ~7% but still big boost I guess. Maybe leaky is part of reason why Mixup can boost a lot on old data, maybe.</p>",
          "rawMarkdown": "After ajust some sequence of my augmentations, now my Mixup can boost LB+0.01, not good as ~7% but still big boost I guess. Maybe leaky is part of reason why Mixup can boost a lot on old data, maybe.",
          "votes": 3
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 1400947,
      "author_name": "CPMP",
      "author_url": "",
      "post_date": "2021-07-26T18:22:51.463000",
      "content": "<p>I didn't remove mixup maybe I should try.</p>",
      "votes": 0,
      "replies": [
        {
          "id": 1401312,
          "author_name": "Hao",
          "author_url": "",
          "post_date": "2021-07-27T07:50:23.380000",
          "content": "<p>After ajust some sequence of my augmentations, now my Mixup can boost LB+0.01, not good as ~7% but still big boost I guess. Maybe leaky is part of reason why Mixup can boost a lot on old data, maybe.</p>",
          "votes": 3,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1400528": "In the history dicussion many people said Mixup is very effective for this competition, like in this [discussion](https://www.kaggle.com/c/seti-breakthrough-listen/discussion/245152)(thanks @ayuraj) said ~7% boost in score, but in my experiment on new data I didn`t notice obvious boost with stanard Mixup with 0.5 probability, has any one observe the same? Shall I make my Mixup more aggressive to 1.0 probability?",
    "1400947": "I didn't remove mixup maybe I should try."
  }
}