{
  "id": 173674,
  "title": "2 Model trained on Class 0 OR Class 1 only",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/173674",
  "author_name": "",
  "post_date": "2020-08-10T08:21:00.301711400Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Here is the link of dataset having model trained on class 0 OR class 1 only.<br>\nModel name having M0 is trained only on class 0 and M1 only on class 1.<br>\nCould we useful in identification of some images.<br>\nHow to use is up to you.<br>\n<a href=\"https://www.kaggle.com/rajnishe/siimoneclassm0m1model\" target=\"_blank\">https://www.kaggle.com/rajnishe/siimoneclassm0m1model</a></p>",
  "messages": [
    {
      "id": "964866",
      "postDate": "08/10/2020 08:21:00",
      "content": "<p>Here is the link of dataset having model trained on class 0 OR class 1 only.<br>\nModel name having M0 is trained only on class 0 and M1 only on class 1.<br>\nCould we useful in identification of some images.<br>\nHow to use is up to you.<br>\n<a href=\"https://www.kaggle.com/rajnishe/siimoneclassm0m1model\" target=\"_blank\">https://www.kaggle.com/rajnishe/siimoneclassm0m1model</a></p>",
      "rawMarkdown": "Here is the link of dataset having model trained on class 0 OR class 1 only.\nModel name having M0 is trained only on class 0 and M1 only on class 1.\nCould we useful in identification of some images.\nHow to use is up to you.\nhttps://www.kaggle.com/rajnishe/siimoneclassm0m1model",
      "votes": null
    },
    {
      "id": "965063",
      "postDate": "08/10/2020 11:08:37",
      "content": "<p>You remind me of outlier detections. 2 classes imbalanced classification is like anomaly detection but with labels.\nWe can test this on oof first by averaging the results of these 2 models. </p>",
      "rawMarkdown": "You remind me of outlier detections. 2 classes imbalanced classification is like anomaly detection but with labels.\nWe can test this on oof first by averaging the results of these 2 models.",
      "votes": null
    },
    {
      "id": "965188",
      "postDate": "08/10/2020 12:53:27",
      "content": "<p>But does the M0 only predicts 0 no matter the input? (Vice versa)</p>",
      "rawMarkdown": "But does the M0 only predicts 0 no matter the input? (Vice versa)",
      "votes": null
    },
    {
      "id": "966104",
      "postDate": "08/11/2020 06:20:19",
      "content": "<p>Here is one prediction on class 0 image of 2 models. I have similar pattern for all data<br>\nSo it means M1 is higher for class 0 and M0 is lower on Class 0 and vice versa.</p>\n<p>M0 - 2.869526e-09    <br>\nM1 -  0.999987</p>",
      "rawMarkdown": "Here is one prediction on class 0 image of 2 models. I have similar pattern for all data\nSo it means M1 is higher for class 0 and M0 is lower on Class 0 and vice versa.\n\nM0 - 2.869526e-09\t\nM1 -  0.999987",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 965063,
      "author_name": "vicioussong",
      "author_url": "",
      "post_date": "08/10/2020 11:08:37",
      "content": "<p>You remind me of outlier detections. 2 classes imbalanced classification is like anomaly detection but with labels.\nWe can test this on oof first by averaging the results of these 2 models. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 965188,
      "author_name": "quandapro",
      "author_url": "",
      "post_date": "08/10/2020 12:53:27",
      "content": "<p>But does the M0 only predicts 0 no matter the input? (Vice versa)</p>",
      "votes": null,
      "replies": [
        {
          "id": 966104,
          "author_name": "rajnishe",
          "author_url": "",
          "post_date": "08/11/2020 06:20:19",
          "content": "<p>Here is one prediction on class 0 image of 2 models. I have similar pattern for all data<br>\nSo it means M1 is higher for class 0 and M0 is lower on Class 0 and vice versa.</p>\n<p>M0 - 2.869526e-09    <br>\nM1 -  0.999987</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "964866": "Here is the link of dataset having model trained on class 0 OR class 1 only.\nModel name having M0 is trained only on class 0 and M1 only on class 1.\nCould we useful in identification of some images.\nHow to use is up to you.\nhttps://www.kaggle.com/rajnishe/siimoneclassm0m1model",
    "965063": "You remind me of outlier detections. 2 classes imbalanced classification is like anomaly detection but with labels.\nWe can test this on oof first by averaging the results of these 2 models.",
    "965188": "But does the M0 only predicts 0 no matter the input? (Vice versa)",
    "966104": "Here is one prediction on class 0 image of 2 models. I have similar pattern for all data\nSo it means M1 is higher for class 0 and M0 is lower on Class 0 and vice versa.\n\nM0 - 2.869526e-09\t\nM1 -  0.999987"
  },
  "source": "meta"
}