{
  "id": 297127,
  "title": "Accuracy metrics - Keras beginner",
  "url": "/competitions/digit-recognizer/discussion/297127",
  "author_name": "Mahmoud Reda",
  "post_date": "2021-12-25T11:30:45.421000",
  "votes": 2,
  "comment_count": 5,
  "views": null,
  "content": "<p>I'm so confused about the Accuracy &amp; the learning curve.<br>\nI use 2 <code>layers =1024</code> with <code>DroupOut = 0.5</code> and metrics:</p>\n<p><code>model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['categorical_accuracy'])</code></p>\n<p>and I get:<br>\n<code>Val Loss = 2.088017225265503</code> &amp; <code>Val Accuracy = 0.09595238417387009</code></p>\n<p>using <code>model.evaluate()</code> but I get about 80% score on the Leaderboard.</p>\n<p><strong>what metrics should I use? and any other suggestions?</strong></p>",
  "messages": [
    {
      "id": 1628891,
      "postDate": "2021-12-25T11:30:45.423Z",
      "content": "<p>I'm so confused about the Accuracy &amp; the learning curve.<br>\nI use 2 <code>layers =1024</code> with <code>DroupOut = 0.5</code> and metrics:</p>\n<p><code>model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['categorical_accuracy'])</code></p>\n<p>and I get:<br>\n<code>Val Loss = 2.088017225265503</code> &amp; <code>Val Accuracy = 0.09595238417387009</code></p>\n<p>using <code>model.evaluate()</code> but I get about 80% score on the Leaderboard.</p>\n<p><strong>what metrics should I use? and any other suggestions?</strong></p>",
      "rawMarkdown": "I'm so confused about the Accuracy & the learning curve.\nI use 2 `layers =1024` with `DroupOut = 0.5` and metrics:\n\n`model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['categorical_accuracy'])`\n\nand I get:\n`Val Loss = 2.088017225265503` & `Val Accuracy = 0.09595238417387009`\n\nusing `model.evaluate()` but I get about 80% score on the Leaderboard.\n\n**what metrics should I use? and any other suggestions?**",
      "votes": 2
    },
    {
      "id": 1631667,
      "postDate": "2021-12-28T17:05:16.003Z",
      "content": "<p>I looked at your notebook and you seemed to have solved the problem. Do you know what was wrong?</p>",
      "rawMarkdown": "I looked at your notebook and you seemed to have solved the problem. Do you know what was wrong?",
      "replies": [
        {
          "id": 1632260,
          "postDate": "2021-12-29T12:35:47.033Z",
          "content": "<p>No, I still have the same problem, I can't get a 90% score again😅</p>",
          "rawMarkdown": "No, I still have the same problem, I can't get a 90% score again😅"
        }
      ]
    },
    {
      "id": 1636980,
      "postDate": "2022-01-03T12:36:31.333Z",
      "rawMarkdown": "",
      "isDeleted": true,
      "replies": [
        {
          "id": 1637377,
          "postDate": "2022-01-03T20:24:16.977Z",
          "content": "<p>yes, I did. thank you.<br>\n&amp; my focus was on the <code>Val Loss</code></p>",
          "rawMarkdown": "yes, I did. thank you.\n& my focus was on the `Val Loss`"
        }
      ]
    },
    {
      "id": 1636901,
      "postDate": "2022-01-03T11:19:22.207Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 1631667,
      "author_name": "Samuel Cortinhas",
      "author_url": "",
      "post_date": "2021-12-28T17:05:16.003000",
      "content": "<p>I looked at your notebook and you seemed to have solved the problem. Do you know what was wrong?</p>",
      "votes": 0,
      "replies": [
        {
          "id": 1632260,
          "author_name": "Mahmoud Reda",
          "author_url": "",
          "post_date": "2021-12-29T12:35:47.033000",
          "content": "<p>No, I still have the same problem, I can't get a 90% score again😅</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1636980,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-01-03T12:36:31.333000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 1637377,
          "author_name": "Mahmoud Reda",
          "author_url": "",
          "post_date": "2022-01-03T20:24:16.977000",
          "content": "<p>yes, I did. thank you.<br>\n&amp; my focus was on the <code>Val Loss</code></p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 1636901,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-01-03T11:19:22.207000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1628891": "I'm so confused about the Accuracy & the learning curve.\nI use 2 `layers =1024` with `DroupOut = 0.5` and metrics:\n\n`model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['categorical_accuracy'])`\n\nand I get:\n`Val Loss = 2.088017225265503` & `Val Accuracy = 0.09595238417387009`\n\nusing `model.evaluate()` but I get about 80% score on the Leaderboard.\n\n**what metrics should I use? and any other suggestions?**",
    "1631667": "I looked at your notebook and you seemed to have solved the problem. Do you know what was wrong?",
    "1636980": "",
    "1636901": ""
  }
}