{
  "id": 12601,
  "title": "Does image order matter",
  "url": "/competitions/datasciencebowl/discussion/12601",
  "author_name": "",
  "post_date": "2015-02-24T17:56:28.507Z",
  "votes": null,
  "comment_count": 4,
  "views": 793,
  "content": "<p>my algorithm only predicts 120898 images. I have correctly provided names of those images and their probability for every class. But i cant do submission unless the excel file has all images. So M planning on appending names of other images at the bottom and assigning some random probability. Would that work?</p>\n<p>Is there any other way. can i just give random 200 images names &nbsp;just to make a total of 130400 images.&nbsp;</p>\n\n<p>thanks</p>",
  "messages": [
    {
      "id": "64813",
      "postDate": "02/24/2015 17:56:28",
      "content": "<p>my algorithm only predicts 120898 images. I have correctly provided names of those images and their probability for every class. But i cant do submission unless the excel file has all images. So M planning on appending names of other images at the bottom and assigning some random probability. Would that work?</p>\n<p>Is there any other way. can i just give random 200 images names &nbsp;just to make a total of 130400 images.&nbsp;</p>\n\n<p>thanks</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64816",
      "postDate": "02/24/2015 18:19:31",
      "content": "<p>You need all 130400 rows or I think you'll get an error. Behind the scenes I think (but could be wrong) that Kaggle first does a join on id between your submission and their solution, so not having all rows is a problem. If your algorithm truly doesn't score all observations, then, as you said, append those Ids at the end of the file and fill in the probabilities with whatever you see fit. That should allow you to at least get your file scored on the LB.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64818",
      "postDate": "02/24/2015 20:08:49",
      "content": "<p>There are 130400 images. I dont know how to find which 200 images are missing from my dataset. I tried to load into sql database using union select but its taking forever. Is it ok if i just randomly name remaining 200 images. or do the names need to match...</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64820",
      "postDate": "02/24/2015 20:30:43",
      "content": "<p>Hmm... that should be fairly straightforward to find. A Python script example that does something similar:</p>\n<p><code>import os</code><code></code></p>\n<p><code>files1 = set(os.listdir('PATH1'))<br>files2 = set(os.listdir('PATH2'))</code></p>\n<p><code>files2.difference(files1)</code></p>\n\n<p>HTH.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "64824",
      "postDate": "02/24/2015 21:07:44",
      "content": "<p>I did a vlookup in excel.. it was time consuming slow but its done for now.. But thanks for that code though. I never knew about os in python.</p>\n<p>thanks</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 64816,
      "author_name": "adjgiulio",
      "author_url": "",
      "post_date": "02/24/2015 18:19:31",
      "content": "<p>You need all 130400 rows or I think you'll get an error. Behind the scenes I think (but could be wrong) that Kaggle first does a join on id between your submission and their solution, so not having all rows is a problem. If your algorithm truly doesn't score all observations, then, as you said, append those Ids at the end of the file and fill in the probabilities with whatever you see fit. That should allow you to at least get your file scored on the LB.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 64818,
      "author_name": "aseem1",
      "author_url": "",
      "post_date": "02/24/2015 20:08:49",
      "content": "<p>There are 130400 images. I dont know how to find which 200 images are missing from my dataset. I tried to load into sql database using union select but its taking forever. Is it ok if i just randomly name remaining 200 images. or do the names need to match...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 64820,
      "author_name": "asaini",
      "author_url": "",
      "post_date": "02/24/2015 20:30:43",
      "content": "<p>Hmm... that should be fairly straightforward to find. A Python script example that does something similar:</p>\n<p><code>import os</code><code></code></p>\n<p><code>files1 = set(os.listdir('PATH1'))<br>files2 = set(os.listdir('PATH2'))</code></p>\n<p><code>files2.difference(files1)</code></p>\n\n<p>HTH.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 64824,
      "author_name": "aseem1",
      "author_url": "",
      "post_date": "02/24/2015 21:07:44",
      "content": "<p>I did a vlookup in excel.. it was time consuming slow but its done for now.. But thanks for that code though. I never knew about os in python.</p>\n<p>thanks</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "64813": "",
    "64816": "",
    "64818": "",
    "64820": "",
    "64824": ""
  },
  "source": "meta"
}