{
  "id": 240179,
  "title": "Anyone using Dacon.ai top solutions.",
  "url": "/competitions/bms-molecular-translation/discussion/240179",
  "author_name": "",
  "post_date": "2021-05-18T18:44:54.866919800Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>This post is about the approach taken by winners in a similar competition by dacon.ai.<br>\nPlease refer to the following discussions for a brief overview:</p>\n<p><a href=\"https://www.kaggle.com/c/bms-molecular-translation/discussion/229978\" target=\"_blank\">https://www.kaggle.com/c/bms-molecular-translation/discussion/229978</a><br>\n<a href=\"https://www.kaggle.com/c/bms-molecular-translation/discussion/229984\" target=\"_blank\">https://www.kaggle.com/c/bms-molecular-translation/discussion/229984</a></p>\n<p>So did anyone try these. How are the results so far. </p>",
  "messages": [
    {
      "id": "1313858",
      "postDate": "05/18/2021 18:44:54",
      "content": "<p>This post is about the approach taken by winners in a similar competition by dacon.ai.<br>\nPlease refer to the following discussions for a brief overview:</p>\n<p><a href=\"https://www.kaggle.com/c/bms-molecular-translation/discussion/229978\" target=\"_blank\">https://www.kaggle.com/c/bms-molecular-translation/discussion/229978</a><br>\n<a href=\"https://www.kaggle.com/c/bms-molecular-translation/discussion/229984\" target=\"_blank\">https://www.kaggle.com/c/bms-molecular-translation/discussion/229984</a></p>\n<p>So did anyone try these. How are the results so far. </p>",
      "rawMarkdown": "This post is about the approach taken by winners in a similar competition by dacon.ai.\nPlease refer to the following discussions for a brief overview:\n\nhttps://www.kaggle.com/c/bms-molecular-translation/discussion/229978\nhttps://www.kaggle.com/c/bms-molecular-translation/discussion/229984\n\nSo did anyone try these. How are the results so far.",
      "votes": null
    },
    {
      "id": "1313866",
      "postDate": "05/18/2021 18:50:08",
      "content": "<p>No, but I borrowed one idea: sample molecules based on some properties.</p>",
      "rawMarkdown": "No, but I borrowed one idea: sample molecules based on some properties.",
      "votes": null
    },
    {
      "id": "1315206",
      "postDate": "05/19/2021 16:17:42",
      "content": "<p>sound interesting. Can you please elaborate a bit.</p>",
      "rawMarkdown": "sound interesting. Can you please elaborate a bit.",
      "votes": null
    },
    {
      "id": "1315342",
      "postDate": "05/19/2021 18:10:59",
      "content": "<p>They only trained on a fraction of all the given molecules based on how difficult they are.</p>\n<p>I train on all of them, but I use PyTorch's WeightedRandomSampler to sample them with different frequency. I add a weight to each molecule based on how many molecules they have, if they have a rare layer, etc.</p>",
      "rawMarkdown": "They only trained on a fraction of all the given molecules based on how difficult they are.\n\nI train on all of them, but I use PyTorch's WeightedRandomSampler to sample them with different frequency. I add a weight to each molecule based on how many molecules they have, if they have a rare layer, etc.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1313866,
      "author_name": "nofreewill",
      "author_url": "",
      "post_date": "05/18/2021 18:50:08",
      "content": "<p>No, but I borrowed one idea: sample molecules based on some properties.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1315206,
          "author_name": "vikrant06",
          "author_url": "",
          "post_date": "05/19/2021 16:17:42",
          "content": "<p>sound interesting. Can you please elaborate a bit.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1315342,
          "author_name": "nofreewill",
          "author_url": "",
          "post_date": "05/19/2021 18:10:59",
          "content": "<p>They only trained on a fraction of all the given molecules based on how difficult they are.</p>\n<p>I train on all of them, but I use PyTorch's WeightedRandomSampler to sample them with different frequency. I add a weight to each molecule based on how many molecules they have, if they have a rare layer, etc.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1313858": "This post is about the approach taken by winners in a similar competition by dacon.ai.\nPlease refer to the following discussions for a brief overview:\n\nhttps://www.kaggle.com/c/bms-molecular-translation/discussion/229978\nhttps://www.kaggle.com/c/bms-molecular-translation/discussion/229984\n\nSo did anyone try these. How are the results so far.",
    "1313866": "No, but I borrowed one idea: sample molecules based on some properties.",
    "1315206": "sound interesting. Can you please elaborate a bit.",
    "1315342": "They only trained on a fraction of all the given molecules based on how difficult they are.\n\nI train on all of them, but I use PyTorch's WeightedRandomSampler to sample them with different frequency. I add a weight to each molecule based on how many molecules they have, if they have a rare layer, etc."
  },
  "source": "meta"
}