{
  "id": 42755,
  "title": "Any loss for only top1 acc?",
  "url": "/competitions/cdiscount-image-classification-challenge/discussion/42755",
  "author_name": "",
  "post_date": "2017-11-04T08:47:24.005829Z",
  "votes": null,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Any loss for only top1 acc?\ndo anybody try the large margin softmax loss</p>",
  "messages": [
    {
      "id": "239707",
      "postDate": "11/04/2017 08:47:24",
      "content": "<p>Any loss for only top1 acc?\ndo anybody try the large margin softmax loss</p>",
      "rawMarkdown": "Any loss for only top1 acc?\ndo anybody try the large margin softmax loss",
      "votes": null
    },
    {
      "id": "239730",
      "postDate": "11/04/2017 10:56:03",
      "content": "<p>I tried the large margin softmax loss but failed to train.\naccuracy is too low after a few iterations.\nI think maybe it was my mistake.</p>",
      "rawMarkdown": "I tried the large margin softmax loss but failed to train.\naccuracy is too low after a few iterations.\nI think maybe it was my mistake.",
      "votes": null
    },
    {
      "id": "239782",
      "postDate": "11/04/2017 15:27:11",
      "content": "<p>now, your performance is single model?</p>",
      "rawMarkdown": "now, your performance is single model?",
      "votes": null
    },
    {
      "id": "239799",
      "postDate": "11/04/2017 16:19:29",
      "content": "<p>no, not a single model.</p>",
      "rawMarkdown": "no, not a single model.",
      "votes": null
    },
    {
      "id": "239920",
      "postDate": "11/05/2017 02:05:46",
      "content": "<p>ok, thanks</p>",
      "rawMarkdown": "ok, thanks",
      "votes": null
    },
    {
      "id": "244323",
      "postDate": "11/16/2017 01:56:35",
      "content": "<p>see this: Loss Functions for Top-k Error: Analysis and Insights\n<a href=\"https://arxiv.org/abs/1512.00486\">https://arxiv.org/abs/1512.00486</a></p>",
      "rawMarkdown": "see this: Loss Functions for Top-k Error: Analysis and Insights\nhttps://arxiv.org/abs/1512.00486",
      "votes": null
    },
    {
      "id": "244956",
      "postDate": "11/17/2017 09:04:19",
      "content": "<p>Have you tried it in a single model? Did it perform better?</p>",
      "rawMarkdown": "Have you tried it in a single model? Did it perform better?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 239730,
      "author_name": "mwbyeon",
      "author_url": "",
      "post_date": "11/04/2017 10:56:03",
      "content": "<p>I tried the large margin softmax loss but failed to train.\naccuracy is too low after a few iterations.\nI think maybe it was my mistake.</p>",
      "votes": null,
      "replies": [
        {
          "id": 239782,
          "author_name": "ifighting",
          "author_url": "",
          "post_date": "11/04/2017 15:27:11",
          "content": "<p>now, your performance is single model?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 239799,
          "author_name": "mwbyeon",
          "author_url": "",
          "post_date": "11/04/2017 16:19:29",
          "content": "<p>no, not a single model.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 239920,
          "author_name": "ifighting",
          "author_url": "",
          "post_date": "11/05/2017 02:05:46",
          "content": "<p>ok, thanks</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 244323,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "11/16/2017 01:56:35",
      "content": "<p>see this: Loss Functions for Top-k Error: Analysis and Insights\n<a href=\"https://arxiv.org/abs/1512.00486\">https://arxiv.org/abs/1512.00486</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 244956,
      "author_name": "zhangsongwei",
      "author_url": "",
      "post_date": "11/17/2017 09:04:19",
      "content": "<p>Have you tried it in a single model? Did it perform better?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "239707": "Any loss for only top1 acc?\ndo anybody try the large margin softmax loss",
    "239730": "I tried the large margin softmax loss but failed to train.\naccuracy is too low after a few iterations.\nI think maybe it was my mistake.",
    "239782": "now, your performance is single model?",
    "239799": "no, not a single model.",
    "239920": "ok, thanks",
    "244323": "see this: Loss Functions for Top-k Error: Analysis and Insights\nhttps://arxiv.org/abs/1512.00486",
    "244956": "Have you tried it in a single model? Did it perform better?"
  },
  "source": "meta"
}