{
  "id": 93148,
  "title": "My Top 30 Features",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/93148",
  "author_name": "",
  "post_date": "2019-05-23T17:37:04.128970200Z",
  "votes": 17,
  "comment_count": 6,
  "views": 0,
  "content": "<p>For a bit of fun here are my top GP Features</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/535959/13280/top30.png\" alt=\"Top 30\"></p>",
  "messages": [
    {
      "id": "535959",
      "postDate": "05/23/2019 17:37:04",
      "content": "<p>For a bit of fun here are my top GP Features</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/535959/13280/top30.png\" alt=\"Top 30\"></p>",
      "rawMarkdown": "For a bit of fun here are my top GP Features\n\n![Top 30](https://storage.googleapis.com/kaggle-forum-message-attachments/535959/13280/top30.png)",
      "votes": null
    },
    {
      "id": "535961",
      "postDate": "05/23/2019 17:41:03",
      "content": "<p>The features come from Andrew's work!</p>",
      "rawMarkdown": "The features come from Andrew's work!",
      "votes": null
    },
    {
      "id": "536663",
      "postDate": "05/25/2019 00:06:07",
      "content": "<p>Thank you for sharing.Can you answer my these two questions.\n1.Does these 30 features help you to end up at 89th place.\n2.Which model do you use like ligbm, xgboost or catboost. </p>",
      "rawMarkdown": "Thank you for sharing.Can you answer my these two questions.\n1.Does these 30 features help you to end up at 89th place.\n2.Which model do you use like ligbm, xgboost or catboost.",
      "votes": null
    },
    {
      "id": "536775",
      "postDate": "05/25/2019 08:46:12",
      "content": "<ol>\n<li>Yes but ignore LB as you will definitely overfit (too small of a sample)</li>\n<li>GP Features == Genetic Programming Features not lightgbm, xgboost or catboost\n(I am using this toy competition to debug my software - I am not interested in medals)</li>\n</ol>",
      "rawMarkdown": "1. Yes but ignore LB as you will definitely overfit (too small of a sample)\n2. GP Features == Genetic Programming Features not lightgbm, xgboost or catboost\n(I am using this toy competition to debug my software - I am not interested in medals)",
      "votes": null
    },
    {
      "id": "537151",
      "postDate": "05/26/2019 09:54:57",
      "content": "<p>&gt; ignore LB as you will definitely overfit (too small of a sample)</p>\n\n<p>You mean the overfiiting to public LB by LB probing, or just hyperparams/validation scheme that can lead to overfitting?</p>\n\n<p>(The question is on your GP kernel really).</p>",
      "rawMarkdown": "&gt; ignore LB as you will definitely overfit (too small of a sample)\n\nYou mean the overfiiting to public LB by LB probing, or just hyperparams/validation scheme that can lead to overfitting?\n\n(The question is on your GP kernel really).",
      "votes": null
    },
    {
      "id": "537222",
      "postDate": "05/26/2019 13:39:27",
      "content": "<p>No LB probing - just instead of early stopping I just stopped at 512 epochs (It's too late to the deadline for me to be any more explicit I am afraid - apologies)</p>",
      "rawMarkdown": "No LB probing - just instead of early stopping I just stopped at 512 epochs (It's too late to the deadline for me to be any more explicit I am afraid - apologies)",
      "votes": null
    },
    {
      "id": "537413",
      "postDate": "05/27/2019 01:11:43",
      "content": "<p>Thanks for sharing.</p>",
      "rawMarkdown": "Thanks for sharing.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 535961,
      "author_name": "scirpus",
      "author_url": "",
      "post_date": "05/23/2019 17:41:03",
      "content": "<p>The features come from Andrew's work!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 536663,
      "author_name": "love123",
      "author_url": "",
      "post_date": "05/25/2019 00:06:07",
      "content": "<p>Thank you for sharing.Can you answer my these two questions.\n1.Does these 30 features help you to end up at 89th place.\n2.Which model do you use like ligbm, xgboost or catboost. </p>",
      "votes": null,
      "replies": [
        {
          "id": 536775,
          "author_name": "scirpus",
          "author_url": "",
          "post_date": "05/25/2019 08:46:12",
          "content": "<ol>\n<li>Yes but ignore LB as you will definitely overfit (too small of a sample)</li>\n<li>GP Features == Genetic Programming Features not lightgbm, xgboost or catboost\n(I am using this toy competition to debug my software - I am not interested in medals)</li>\n</ol>",
          "votes": null,
          "replies": []
        },
        {
          "id": 537151,
          "author_name": "sergeyzlobin",
          "author_url": "",
          "post_date": "05/26/2019 09:54:57",
          "content": "<p>&gt; ignore LB as you will definitely overfit (too small of a sample)</p>\n\n<p>You mean the overfiiting to public LB by LB probing, or just hyperparams/validation scheme that can lead to overfitting?</p>\n\n<p>(The question is on your GP kernel really).</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 537222,
          "author_name": "scirpus",
          "author_url": "",
          "post_date": "05/26/2019 13:39:27",
          "content": "<p>No LB probing - just instead of early stopping I just stopped at 512 epochs (It's too late to the deadline for me to be any more explicit I am afraid - apologies)</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 537413,
      "author_name": "longyin2",
      "author_url": "",
      "post_date": "05/27/2019 01:11:43",
      "content": "<p>Thanks for sharing.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "535959": "For a bit of fun here are my top GP Features\n\n![Top 30](https://storage.googleapis.com/kaggle-forum-message-attachments/535959/13280/top30.png)",
    "535961": "The features come from Andrew's work!",
    "536663": "Thank you for sharing.Can you answer my these two questions.\n1.Does these 30 features help you to end up at 89th place.\n2.Which model do you use like ligbm, xgboost or catboost.",
    "536775": "1. Yes but ignore LB as you will definitely overfit (too small of a sample)\n2. GP Features == Genetic Programming Features not lightgbm, xgboost or catboost\n(I am using this toy competition to debug my software - I am not interested in medals)",
    "537151": "&gt; ignore LB as you will definitely overfit (too small of a sample)\n\nYou mean the overfiiting to public LB by LB probing, or just hyperparams/validation scheme that can lead to overfitting?\n\n(The question is on your GP kernel really).",
    "537222": "No LB probing - just instead of early stopping I just stopped at 512 epochs (It's too late to the deadline for me to be any more explicit I am afraid - apologies)",
    "537413": "Thanks for sharing."
  },
  "source": "meta"
}