{
  "id": 78827,
  "title": "When your local CV gets 0.695+",
  "url": "/competitions/quora-insincere-questions-classification/discussion/78827",
  "author_name": "spongebob",
  "post_date": "2019-01-28T09:18:44.154000",
  "votes": 7,
  "comment_count": 9,
  "views": 0,
  "content": "<p>1.How about the lb??</p>\n\n<p>2.What do you think about this??</p>\n\n<p>3.Variance of test preds in 5 k-folds does really matter??</p>\n\n<p>4.Trust your local CV??</p>",
  "messages": [
    {
      "id": 462422,
      "postDate": "2019-01-28T09:18:44.153Z",
      "content": "<p>1.How about the lb??</p>\n\n<p>2.What do you think about this??</p>\n\n<p>3.Variance of test preds in 5 k-folds does really matter??</p>\n\n<p>4.Trust your local CV??</p>",
      "rawMarkdown": "1.How about the lb??\n\n2.What do you think about this??\n\n3.Variance of test preds in 5 k-folds does really matter??\n\n4.Trust your local CV??",
      "votes": 7
    },
    {
      "id": 462779,
      "postDate": "2019-01-28T21:52:20.127Z",
      "content": "<p>You should use proper neural net , proper weights and epoch size.\nMay be u can get that by different word em belling.\nUse different strategy and you will get output better.</p>",
      "rawMarkdown": "You should use proper neural net , proper weights and epoch size.\nMay be u can get that by different word em belling.\nUse different strategy and you will get output better.",
      "votes": 8
    },
    {
      "id": 462678,
      "postDate": "2019-01-28T17:55:41.377Z",
      "content": "<p>When your local CV gets 0.695+. When? :(</p>",
      "rawMarkdown": "When your local CV gets 0.695+. When? :(",
      "votes": 7
    },
    {
      "id": 462440,
      "postDate": "2019-01-28T09:48:22.193Z",
      "content": "<p>My 4 fold local CV is about 0.699-0.701 . In kernel I use no fold due to time limit. \n1. lb vary between 0.691 to 0.701\n2. randomness, test set is too small, and something else...\n3. ...\n4. I don't konw... At last I probably choose one highest local CV and one highest LB.</p>",
      "rawMarkdown": "My 4 fold local CV is about 0.699-0.701 . In kernel I use no fold due to time limit. \n1. lb vary between 0.691 to 0.701\n2. randomness, test set is too small, and something else...\n3. ...\n4. I don't konw... At last I probably choose one highest local CV and one highest LB.",
      "votes": 5,
      "replies": [
        {
          "id": 464000,
          "postDate": "2019-01-31T03:02:06.347Z",
          "content": "<p>Why is this downvoted I don't understand zzz</p>",
          "rawMarkdown": "Why is this downvoted I don't understand zzz",
          "votes": 1
        },
        {
          "id": 464059,
          "postDate": "2019-01-31T06:10:53.613Z",
          "content": "<p>maybe randomness is really strong，haha</p>",
          "rawMarkdown": "maybe randomness is really strong，haha",
          "votes": 1
        }
      ]
    },
    {
      "id": 462542,
      "postDate": "2019-01-28T13:26:34.597Z",
      "content": "<p><code>??</code> == <code>sincere</code>  (A-A</p>",
      "rawMarkdown": "`??` == `sincere`  (A-A",
      "replies": [
        {
          "id": 462618,
          "postDate": "2019-01-28T15:47:06.780Z",
          "content": "<p>miao miao miao （down votes happen to this. So it is insincere, 2333）</p>",
          "rawMarkdown": "miao miao miao （down votes happen to this. So it is insincere, 2333）",
          "votes": -4
        }
      ]
    },
    {
      "id": 464599,
      "postDate": "2019-02-01T06:23:43.923Z",
      "content": "<p>I had a local CV of 0.72+ after stacking as well as simple averaging. But, my public LB was 0.699. I have no idea, should I go with that or not.</p>",
      "rawMarkdown": "I had a local CV of 0.72+ after stacking as well as simple averaging. But, my public LB was 0.699. I have no idea, should I go with that or not."
    },
    {
      "id": 462778,
      "postDate": "2019-01-28T21:50:21.560Z",
      "rawMarkdown": "",
      "votes": 1,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 462779,
      "author_name": "siddharth",
      "author_url": "",
      "post_date": "2019-01-28T21:52:20.127000",
      "content": "<p>You should use proper neural net , proper weights and epoch size.\nMay be u can get that by different word em belling.\nUse different strategy and you will get output better.</p>",
      "votes": 8,
      "replies": []
    },
    {
      "id": 462678,
      "author_name": "Oleg Yaroshevskiy",
      "author_url": "",
      "post_date": "2019-01-28T17:55:41.377000",
      "content": "<p>When your local CV gets 0.695+. When? :(</p>",
      "votes": 7,
      "replies": []
    },
    {
      "id": 462440,
      "author_name": "HZD",
      "author_url": "",
      "post_date": "2019-01-28T09:48:22.193000",
      "content": "<p>My 4 fold local CV is about 0.699-0.701 . In kernel I use no fold due to time limit. \n1. lb vary between 0.691 to 0.701\n2. randomness, test set is too small, and something else...\n3. ...\n4. I don't konw... At last I probably choose one highest local CV and one highest LB.</p>",
      "votes": 5,
      "replies": [
        {
          "id": 464000,
          "author_name": "Mekaveli.",
          "author_url": "",
          "post_date": "2019-01-31T03:02:06.347000",
          "content": "<p>Why is this downvoted I don't understand zzz</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 464059,
          "author_name": "HZD",
          "author_url": "",
          "post_date": "2019-01-31T06:10:53.613000",
          "content": "<p>maybe randomness is really strong，haha</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 462542,
      "author_name": "Hikkiiiiiiiii",
      "author_url": "",
      "post_date": "2019-01-28T13:26:34.597000",
      "content": "<p><code>??</code> == <code>sincere</code>  (A-A</p>",
      "votes": 0,
      "replies": [
        {
          "id": 462618,
          "author_name": "spongebob",
          "author_url": "",
          "post_date": "2019-01-28T15:47:06.780000",
          "content": "<p>miao miao miao （down votes happen to this. So it is insincere, 2333）</p>",
          "votes": -4,
          "replies": []
        }
      ]
    },
    {
      "id": 464599,
      "author_name": "aintnosunshine",
      "author_url": "",
      "post_date": "2019-02-01T06:23:43.923000",
      "content": "<p>I had a local CV of 0.72+ after stacking as well as simple averaging. But, my public LB was 0.699. I have no idea, should I go with that or not.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 462778,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-01-28T21:50:21.560000",
      "content": "",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "462422": "1.How about the lb??\n\n2.What do you think about this??\n\n3.Variance of test preds in 5 k-folds does really matter??\n\n4.Trust your local CV??",
    "462779": "You should use proper neural net , proper weights and epoch size.\nMay be u can get that by different word em belling.\nUse different strategy and you will get output better.",
    "462678": "When your local CV gets 0.695+. When? :(",
    "462440": "My 4 fold local CV is about 0.699-0.701 . In kernel I use no fold due to time limit. \n1. lb vary between 0.691 to 0.701\n2. randomness, test set is too small, and something else...\n3. ...\n4. I don't konw... At last I probably choose one highest local CV and one highest LB.",
    "462542": "`??` == `sincere`  (A-A",
    "464599": "I had a local CV of 0.72+ after stacking as well as simple averaging. But, my public LB was 0.699. I have no idea, should I go with that or not.",
    "462778": ""
  }
}