{
  "id": 193341,
  "title": "Is There a Way to Train LightGBM using Data Table ?",
  "url": "/competitions/riiid-test-answer-prediction/discussion/193341",
  "author_name": "",
  "post_date": "2020-10-26T16:11:04.275711100Z",
  "votes": 3,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I just tried out FTRL , with Data table kernel provided by <a href=\"https://www.kaggle.com/rohanrao\" target=\"_blank\">@rohanrao</a> <a href=\"https://www.kaggle.com/rohanrao/riiid-ftrl-ftw\" target=\"_blank\">here</a> , I have seen it works phenomenally well with data table ! When it comes to Xgboost or Lightgbm people resort to using pandas dataframe with some sub sampling and also it limits amount of Data and Features we can use as it eats up lot of RAM ! So I wanted to know if anybody knows whether there is a way to use Data Table along with Lightgbm , if there is we might be able to exploit Lightgbm or Xgboost more nicely as they work really well for Tabular Data ! </p>",
  "messages": [
    {
      "id": "1060921",
      "postDate": "10/26/2020 16:11:04",
      "content": "<p>I just tried out FTRL , with Data table kernel provided by <a href=\"https://www.kaggle.com/rohanrao\" target=\"_blank\">@rohanrao</a> <a href=\"https://www.kaggle.com/rohanrao/riiid-ftrl-ftw\" target=\"_blank\">here</a> , I have seen it works phenomenally well with data table ! When it comes to Xgboost or Lightgbm people resort to using pandas dataframe with some sub sampling and also it limits amount of Data and Features we can use as it eats up lot of RAM ! So I wanted to know if anybody knows whether there is a way to use Data Table along with Lightgbm , if there is we might be able to exploit Lightgbm or Xgboost more nicely as they work really well for Tabular Data ! </p>",
      "rawMarkdown": "I just tried out FTRL , with Data table kernel provided by @rohanrao [here](https://www.kaggle.com/rohanrao/riiid-ftrl-ftw) , I have seen it works phenomenally well with data table ! When it comes to Xgboost or Lightgbm people resort to using pandas dataframe with some sub sampling and also it limits amount of Data and Features we can use as it eats up lot of RAM ! So I wanted to know if anybody knows whether there is a way to use Data Table along with Lightgbm , if there is we might be able to exploit Lightgbm or Xgboost more nicely as they work really well for Tabular Data !",
      "votes": null
    },
    {
      "id": "1060927",
      "postDate": "10/26/2020 16:16:30",
      "content": "<p>Many folks have mentioned <a href=\"https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/192919\" target=\"_blank\">here</a> that using a sample of data is good enough to get a competitive score.</p>",
      "rawMarkdown": "Many folks have mentioned [here](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/192919) that using a sample of data is good enough to get a competitive score.",
      "votes": null
    },
    {
      "id": "1060933",
      "postDate": "10/26/2020 16:22:57",
      "content": "<p><a href=\"https://www.kaggle.com/vopani\" target=\"_blank\">@vopani</a> Thanks !</p>",
      "rawMarkdown": "vopani Thanks !",
      "votes": null
    },
    {
      "id": "1061111",
      "postDate": "10/26/2020 18:49:09",
      "content": "<p>What if the goal is to be <code>not just good enough</code> 😄</p>",
      "rawMarkdown": "What if the goal is to be `not just good enough` 😄",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1060927,
      "author_name": "rohanrao",
      "author_url": "",
      "post_date": "10/26/2020 16:16:30",
      "content": "<p>Many folks have mentioned <a href=\"https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/192919\" target=\"_blank\">here</a> that using a sample of data is good enough to get a competitive score.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1060933,
          "author_name": "sayedathar11",
          "author_url": "",
          "post_date": "10/26/2020 16:22:57",
          "content": "<p><a href=\"https://www.kaggle.com/vopani\" target=\"_blank\">@vopani</a> Thanks !</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1061111,
          "author_name": "yihdarshieh",
          "author_url": "",
          "post_date": "10/26/2020 18:49:09",
          "content": "<p>What if the goal is to be <code>not just good enough</code> 😄</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1060921": "I just tried out FTRL , with Data table kernel provided by @rohanrao [here](https://www.kaggle.com/rohanrao/riiid-ftrl-ftw) , I have seen it works phenomenally well with data table ! When it comes to Xgboost or Lightgbm people resort to using pandas dataframe with some sub sampling and also it limits amount of Data and Features we can use as it eats up lot of RAM ! So I wanted to know if anybody knows whether there is a way to use Data Table along with Lightgbm , if there is we might be able to exploit Lightgbm or Xgboost more nicely as they work really well for Tabular Data !",
    "1060927": "Many folks have mentioned [here](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/192919) that using a sample of data is good enough to get a competitive score.",
    "1060933": "vopani Thanks !",
    "1061111": "What if the goal is to be `not just good enough` 😄"
  },
  "source": "meta"
}