{
  "id": 98561,
  "title": "pairs of atoms in train/test data",
  "url": "/competitions/champs-scalar-coupling/discussion/98561",
  "author_name": "",
  "post_date": "2019-07-04T15:40:40.850128800Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "<p>When we look at first molecule:</p>\n\n<p>dsgdb9nsd_000001</p>\n\n<p>it has 5 atoms in structures\nit means that it is possible to construct  5 * 4 or 5 * 4 / 2 pairs\nin train for this molecule we see 10 rows</p>\n\n<ul>\n<li>for molecule dsgdb9nsd_000003 we have 3 atoms in structures and 1 row in data</li>\n<li>for molecule dsgdb9nsd_000005 we have 3 atoms in structures and 2 rows in data</li>\n<li>dsgdb9nsd_000008 6 atoms and 10 rows</li>\n<li>dsgdb9nsd_000010 6 atoms and 12 rows</li>\n</ul>\n\n<p>does it mean that only subset of pairs is stored in train/test rows? because kernels use this data to find closest pair and this way some pairs are skipped</p>",
  "messages": [
    {
      "id": "568247",
      "postDate": "07/04/2019 15:40:40",
      "content": "<p>When we look at first molecule:</p>\n\n<p>dsgdb9nsd_000001</p>\n\n<p>it has 5 atoms in structures\nit means that it is possible to construct  5 * 4 or 5 * 4 / 2 pairs\nin train for this molecule we see 10 rows</p>\n\n<ul>\n<li>for molecule dsgdb9nsd_000003 we have 3 atoms in structures and 1 row in data</li>\n<li>for molecule dsgdb9nsd_000005 we have 3 atoms in structures and 2 rows in data</li>\n<li>dsgdb9nsd_000008 6 atoms and 10 rows</li>\n<li>dsgdb9nsd_000010 6 atoms and 12 rows</li>\n</ul>\n\n<p>does it mean that only subset of pairs is stored in train/test rows? because kernels use this data to find closest pair and this way some pairs are skipped</p>",
      "rawMarkdown": "When we look at first molecule:\n\ndsgdb9nsd_000001\n\nit has 5 atoms in structures\nit means that it is possible to construct  5 * 4 or 5 * 4 / 2 pairs\nin train for this molecule we see 10 rows\n\n- for molecule dsgdb9nsd_000003 we have 3 atoms in structures and 1 row in data\n- for molecule dsgdb9nsd_000005 we have 3 atoms in structures and 2 rows in data\n- dsgdb9nsd_000008 6 atoms and 10 rows\n- dsgdb9nsd_000010 6 atoms and 12 rows\n\ndoes it mean that only subset of pairs is stored in train/test rows? because kernels use this data to find closest pair and this way some pairs are skipped",
      "votes": null
    },
    {
      "id": "568259",
      "postDate": "07/04/2019 15:54:44",
      "content": "<p><a href=\"https://www.kaggle.com/c/champs-scalar-coupling/data\">Data Description</a>\n\"For this competition, you will not be predicting <em>all</em> the atom pairs in each molecule rather, you will only need to predict the pairs that are explicitly listed in the train and test files. For example, some molecules contain Fluorine (F), but you will not be predicting the scalar coupling constant for any pair that includes F.\"</p>",
      "rawMarkdown": "[Data Description](https://www.kaggle.com/c/champs-scalar-coupling/data)\n\"For this competition, you will not be predicting *all* the atom pairs in each molecule rather, you will only need to predict the pairs that are explicitly listed in the train and test files. For example, some molecules contain Fluorine (F), but you will not be predicting the scalar coupling constant for any pair that includes F.\"",
      "votes": null
    },
    {
      "id": "568262",
      "postDate": "07/04/2019 15:56:57",
      "content": "<p>First of all: not all atoms are connected with each other, some of them can be connected to only one other aton.</p>\n\n<p>And as alvor pointed out, not all possible pairs are presented in train/test data.</p>",
      "rawMarkdown": "First of all: not all atoms are connected with each other, some of them can be connected to only one other aton.\n\nAnd as alvor pointed out, not all possible pairs are presented in train/test data.",
      "votes": null
    },
    {
      "id": "568263",
      "postDate": "07/04/2019 15:58:19",
      "content": "<p>thanks this is probably that</p>",
      "rawMarkdown": "thanks this is probably that",
      "votes": null
    },
    {
      "id": "568266",
      "postDate": "07/04/2019 15:59:57",
      "content": "<p>In kernels there is algorithm to find min distance and max distance in each molecule, but only for rows presented in train/test data. I was thinking should we check also all possible combinations or just rows.</p>",
      "rawMarkdown": "In kernels there is algorithm to find min distance and max distance in each molecule, but only for rows presented in train/test data. I was thinking should we check also all possible combinations or just rows.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 568259,
      "author_name": "allvor",
      "author_url": "",
      "post_date": "07/04/2019 15:54:44",
      "content": "<p><a href=\"https://www.kaggle.com/c/champs-scalar-coupling/data\">Data Description</a>\n\"For this competition, you will not be predicting <em>all</em> the atom pairs in each molecule rather, you will only need to predict the pairs that are explicitly listed in the train and test files. For example, some molecules contain Fluorine (F), but you will not be predicting the scalar coupling constant for any pair that includes F.\"</p>",
      "votes": null,
      "replies": [
        {
          "id": 568263,
          "author_name": "jacekpoplawski",
          "author_url": "",
          "post_date": "07/04/2019 15:58:19",
          "content": "<p>thanks this is probably that</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 568262,
      "author_name": "artgor",
      "author_url": "",
      "post_date": "07/04/2019 15:56:57",
      "content": "<p>First of all: not all atoms are connected with each other, some of them can be connected to only one other aton.</p>\n\n<p>And as alvor pointed out, not all possible pairs are presented in train/test data.</p>",
      "votes": null,
      "replies": [
        {
          "id": 568266,
          "author_name": "jacekpoplawski",
          "author_url": "",
          "post_date": "07/04/2019 15:59:57",
          "content": "<p>In kernels there is algorithm to find min distance and max distance in each molecule, but only for rows presented in train/test data. I was thinking should we check also all possible combinations or just rows.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "568247": "When we look at first molecule:\n\ndsgdb9nsd_000001\n\nit has 5 atoms in structures\nit means that it is possible to construct  5 * 4 or 5 * 4 / 2 pairs\nin train for this molecule we see 10 rows\n\n- for molecule dsgdb9nsd_000003 we have 3 atoms in structures and 1 row in data\n- for molecule dsgdb9nsd_000005 we have 3 atoms in structures and 2 rows in data\n- dsgdb9nsd_000008 6 atoms and 10 rows\n- dsgdb9nsd_000010 6 atoms and 12 rows\n\ndoes it mean that only subset of pairs is stored in train/test rows? because kernels use this data to find closest pair and this way some pairs are skipped",
    "568259": "[Data Description](https://www.kaggle.com/c/champs-scalar-coupling/data)\n\"For this competition, you will not be predicting *all* the atom pairs in each molecule rather, you will only need to predict the pairs that are explicitly listed in the train and test files. For example, some molecules contain Fluorine (F), but you will not be predicting the scalar coupling constant for any pair that includes F.\"",
    "568262": "First of all: not all atoms are connected with each other, some of them can be connected to only one other aton.\n\nAnd as alvor pointed out, not all possible pairs are presented in train/test data.",
    "568263": "thanks this is probably that",
    "568266": "In kernels there is algorithm to find min distance and max distance in each molecule, but only for rows presented in train/test data. I was thinking should we check also all possible combinations or just rows."
  },
  "source": "meta"
}