{
  "id": 293537,
  "title": "PCA Feature Selection",
  "url": "/competitions/cs5785-fall-2021-final/discussion/293537",
  "author_name": "",
  "post_date": "2021-12-06T00:50:31.223844500Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>As a heads up, it seems that PCA for the features (X values) alone didn't do much for our team: the highest component only explained 17% of the variance, and the second best component explained 10%. </p>",
  "messages": [
    {
      "id": "1607539",
      "postDate": "12/06/2021 00:50:31",
      "content": "<p>As a heads up, it seems that PCA for the features (X values) alone didn't do much for our team: the highest component only explained 17% of the variance, and the second best component explained 10%. </p>",
      "rawMarkdown": "As a heads up, it seems that PCA for the features (X values) alone didn't do much for our team: the highest component only explained 17% of the variance, and the second best component explained 10%.",
      "votes": null
    },
    {
      "id": "1617202",
      "postDate": "12/14/2021 00:06:08",
      "content": "<p>My team saw about 3% improvement with PCA. <br>\nWe also tried using the intermediate Resnet features vs the final ones and saw a small improvement on the intermediate features.</p>",
      "rawMarkdown": "My team saw about 3% improvement with PCA. \nWe also tried using the intermediate Resnet features vs the final ones and saw a small improvement on the intermediate features.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1617202,
      "author_name": "jennyjuarez",
      "author_url": "",
      "post_date": "12/14/2021 00:06:08",
      "content": "<p>My team saw about 3% improvement with PCA. <br>\nWe also tried using the intermediate Resnet features vs the final ones and saw a small improvement on the intermediate features.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1607539": "As a heads up, it seems that PCA for the features (X values) alone didn't do much for our team: the highest component only explained 17% of the variance, and the second best component explained 10%.",
    "1617202": "My team saw about 3% improvement with PCA. \nWe also tried using the intermediate Resnet features vs the final ones and saw a small improvement on the intermediate features."
  },
  "source": "meta"
}