{
  "id": 134094,
  "title": "Tuning parameter by 'lightgbm_tuner'",
  "url": "/competitions/cat-in-the-dat-ii/discussion/134094",
  "author_name": "",
  "post_date": "2020-03-06T00:37:11.071125700Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hello. I learned for the first time a revolutionary method of automatically determining parameters in LGBM. So I tried to use 'lightgbm_tuner' from 'optuna.integration'. After Target Encoding, some parameters were output by this method : 'such as 'num_leaves', 'min_child_samples'.\nHowever, I wondered there are some parameters not specified such as 'learning_rate', 'max_depth'. I think these parameters are very important for good prediction.\nIn conclusion, there are two things I want to be told:\n1.Why some parameters are not specified?\n2.How to determine these parameaters?\nThank you!</p>",
  "messages": [
    {
      "id": "764829",
      "postDate": "03/06/2020 00:37:11",
      "content": "<p>Hello. I learned for the first time a revolutionary method of automatically determining parameters in LGBM. So I tried to use 'lightgbm_tuner' from 'optuna.integration'. After Target Encoding, some parameters were output by this method : 'such as 'num_leaves', 'min_child_samples'.\nHowever, I wondered there are some parameters not specified such as 'learning_rate', 'max_depth'. I think these parameters are very important for good prediction.\nIn conclusion, there are two things I want to be told:\n1.Why some parameters are not specified?\n2.How to determine these parameaters?\nThank you!</p>",
      "rawMarkdown": "Hello. I learned for the first time a revolutionary method of automatically determining parameters in LGBM. So I tried to use 'lightgbm_tuner' from 'optuna.integration'. After Target Encoding, some parameters were output by this method : 'such as 'num_leaves', 'min_child_samples'.\nHowever, I wondered there are some parameters not specified such as 'learning_rate', 'max_depth'. I think these parameters are very important for good prediction.\nIn conclusion, there are two things I want to be told:\n1.Why some parameters are not specified?\n2.How to determine these parameaters?\nThank you!",
      "votes": null
    },
    {
      "id": "764858",
      "postDate": "03/06/2020 01:44:49",
      "content": "<p>In reference, the value of the output parameter.(train_test_split(test_size=0.2, random_state=0)</p>\n\n<p>{'lambda _l1':9.0320008814508263, 'lambda _l2':1.9574345401644684e-6, 'num _leaves':3, 'feature _fraction':0.62, 'bagging _fraction':1.0, 'bagging _freq':0, 'min _child _samples':50}</p>",
      "rawMarkdown": "In reference, the value of the output parameter.(train_test_split(test_size=0.2, random_state=0)\n\n{'lambda _l1':9.0320008814508263, 'lambda _l2':1.9574345401644684e-6, 'num _leaves':3, 'feature _fraction':0.62, 'bagging _fraction':1.0, 'bagging _freq':0, 'min _child _samples':50}",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 764858,
      "author_name": "yutoohnuki",
      "author_url": "",
      "post_date": "03/06/2020 01:44:49",
      "content": "<p>In reference, the value of the output parameter.(train_test_split(test_size=0.2, random_state=0)</p>\n\n<p>{'lambda _l1':9.0320008814508263, 'lambda _l2':1.9574345401644684e-6, 'num _leaves':3, 'feature _fraction':0.62, 'bagging _fraction':1.0, 'bagging _freq':0, 'min _child _samples':50}</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "764829": "Hello. I learned for the first time a revolutionary method of automatically determining parameters in LGBM. So I tried to use 'lightgbm_tuner' from 'optuna.integration'. After Target Encoding, some parameters were output by this method : 'such as 'num_leaves', 'min_child_samples'.\nHowever, I wondered there are some parameters not specified such as 'learning_rate', 'max_depth'. I think these parameters are very important for good prediction.\nIn conclusion, there are two things I want to be told:\n1.Why some parameters are not specified?\n2.How to determine these parameaters?\nThank you!",
    "764858": "In reference, the value of the output parameter.(train_test_split(test_size=0.2, random_state=0)\n\n{'lambda _l1':9.0320008814508263, 'lambda _l2':1.9574345401644684e-6, 'num _leaves':3, 'feature _fraction':0.62, 'bagging _fraction':1.0, 'bagging _freq':0, 'min _child _samples':50}"
  },
  "source": "meta"
}