{
  "id": 63066,
  "title": "Final submission?",
  "url": "/competitions/trackml-particle-identification/discussion/63066",
  "author_name": "",
  "post_date": "2018-08-11T09:56:03.942600600Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Just wondering, at this point of the contest, which team has a local submission that is way higher than their LB and is simply waiting for the last day to submit, and who is still building one that might result in a big improvement? We are still struggling to find a good way to ensemble multiple submissions to yield a higher LB (got the current score from 3 stronger models and a weaker one)</p>",
  "messages": [
    {
      "id": "368923",
      "postDate": "08/11/2018 09:56:03",
      "content": "<p>Just wondering, at this point of the contest, which team has a local submission that is way higher than their LB and is simply waiting for the last day to submit, and who is still building one that might result in a big improvement? We are still struggling to find a good way to ensemble multiple submissions to yield a higher LB (got the current score from 3 stronger models and a weaker one)</p>",
      "rawMarkdown": "Just wondering, at this point of the contest, which team has a local submission that is way higher than their LB and is simply waiting for the last day to submit, and who is still building one that might result in a big improvement? We are still struggling to find a good way to ensemble multiple submissions to yield a higher LB (got the current score from 3 stronger models and a weaker one)",
      "votes": null
    },
    {
      "id": "368935",
      "postDate": "08/11/2018 10:56:49",
      "content": "<p>See <a href=\"https://www.kaggle.com/c/trackml-particle-identification/discussion/62163#latest-367579\">https://www.kaggle.com/c/trackml-particle-identification/discussion/62163#latest-367579</a></p>",
      "rawMarkdown": "See https://www.kaggle.com/c/trackml-particle-identification/discussion/62163#latest-367579",
      "votes": null
    },
    {
      "id": "368985",
      "postDate": "08/11/2018 14:25:00",
      "content": "<p>thanks!</p>",
      "rawMarkdown": "thanks!",
      "votes": null
    },
    {
      "id": "368989",
      "postDate": "08/11/2018 14:31:05",
      "content": "<blockquote>\n  <p>We are still struggling to find a good way to ensemble multiple submissions to yield a higher LB </p>\n</blockquote>\n\n<p>For what it is worth, I am struggling as well for that.</p>\n\n<p>I do find quite a good proportion of the non centered tracks now with DBSCAN, but haven't found yet a way to add them effectively to the centered tracks I have.  I only get a modest improvement so far.</p>",
      "rawMarkdown": "&gt; We are still struggling to find a good way to ensemble multiple submissions to yield a higher LB \n\nFor what it is worth, I am struggling as well for that.\n\nI do find quite a good proportion of the non centered tracks now with DBSCAN, but haven't found yet a way to add them effectively to the centered tracks I have.  I only get a modest improvement so far.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 368935,
      "author_name": "cpmpml",
      "author_url": "",
      "post_date": "08/11/2018 10:56:49",
      "content": "<p>See <a href=\"https://www.kaggle.com/c/trackml-particle-identification/discussion/62163#latest-367579\">https://www.kaggle.com/c/trackml-particle-identification/discussion/62163#latest-367579</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 368985,
          "author_name": "alexanderliao",
          "author_url": "",
          "post_date": "08/11/2018 14:25:00",
          "content": "<p>thanks!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 368989,
          "author_name": "cpmpml",
          "author_url": "",
          "post_date": "08/11/2018 14:31:05",
          "content": "<blockquote>\n  <p>We are still struggling to find a good way to ensemble multiple submissions to yield a higher LB </p>\n</blockquote>\n\n<p>For what it is worth, I am struggling as well for that.</p>\n\n<p>I do find quite a good proportion of the non centered tracks now with DBSCAN, but haven't found yet a way to add them effectively to the centered tracks I have.  I only get a modest improvement so far.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "368923": "Just wondering, at this point of the contest, which team has a local submission that is way higher than their LB and is simply waiting for the last day to submit, and who is still building one that might result in a big improvement? We are still struggling to find a good way to ensemble multiple submissions to yield a higher LB (got the current score from 3 stronger models and a weaker one)",
    "368935": "See https://www.kaggle.com/c/trackml-particle-identification/discussion/62163#latest-367579",
    "368985": "thanks!",
    "368989": "&gt; We are still struggling to find a good way to ensemble multiple submissions to yield a higher LB \n\nFor what it is worth, I am struggling as well for that.\n\nI do find quite a good proportion of the non centered tracks now with DBSCAN, but haven't found yet a way to add them effectively to the centered tracks I have.  I only get a modest improvement so far."
  },
  "source": "meta"
}