{
  "id": 209614,
  "title": "Congratulations and looking forward to learn from top winners' 0.810+ nn models",
  "url": "/competitions/riiid-test-answer-prediction/discussion/209614",
  "author_name": "Daniels",
  "post_date": "2021-01-08T03:09:53.838000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>It is a very exciting competition, I thought the edge of gold medal is LB 0.810 a week ago，but finally I was wrong.</p>\n<p>In past time series competitions, gbdt with feature engineering usually performs well, but the performance gap between gbdt and saint in this competition is obvious. Very surprised to find that the seq2seq model has such an amazing performance on time series problems. It is very worth learning. Congratulations to the top winners and hope to learn from your code and solution sharing！</p>",
  "messages": [
    {
      "id": 1143729,
      "postDate": "2021-01-08T03:09:53.840Z",
      "content": "<p>It is a very exciting competition, I thought the edge of gold medal is LB 0.810 a week ago，but finally I was wrong.</p>\n<p>In past time series competitions, gbdt with feature engineering usually performs well, but the performance gap between gbdt and saint in this competition is obvious. Very surprised to find that the seq2seq model has such an amazing performance on time series problems. It is very worth learning. Congratulations to the top winners and hope to learn from your code and solution sharing！</p>",
      "rawMarkdown": "It is a very exciting competition, I thought the edge of gold medal is LB 0.810 a week ago，but finally I was wrong.\n\nIn past time series competitions, gbdt with feature engineering usually performs well, but the performance gap between gbdt and saint in this competition is obvious. Very surprised to find that the seq2seq model has such an amazing performance on time series problems. It is very worth learning. Congratulations to the top winners and hope to learn from your code and solution sharing！",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1143729": "It is a very exciting competition, I thought the edge of gold medal is LB 0.810 a week ago，but finally I was wrong.\n\nIn past time series competitions, gbdt with feature engineering usually performs well, but the performance gap between gbdt and saint in this competition is obvious. Very surprised to find that the seq2seq model has such an amazing performance on time series problems. It is very worth learning. Congratulations to the top winners and hope to learn from your code and solution sharing！"
  }
}