{
  "id": 583113,
  "title": "How does this happen? Awesome training accuracy/loss! Horrendous validation accuracy/loss...",
  "url": "/competitions/cmi-detect-behavior-with-sensor-data/discussion/583113",
  "author_name": "",
  "post_date": "2025-06-04T19:49:54.132128700Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I successfully created my first draft model for this competition, but I get some interesting behavior that I haven't seen before, nor do I understand what it means and how to solve it.</p>\n<p>While training my model (20 epochs), my training accuracy and loss are impeccable. They look beautiful in my graph below. However, my validation accuracy and loss make me want to vomit.</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F16391251%2Fe5b091e093bc4a959b4ef8f97347059b%2Fbad_validation.png?generation=1749066442410405&amp;alt=media\" alt=\"\"></p>\n<p>I compiled my model with the following line:</p>\n<pre><code>model.compile(=, =, metrics=[])\n</code></pre>\n<p>What are some potential causes for this behavior? What did I overlook to cause such a hot steamy pile of garbage?</p>",
  "messages": [
    {
      "id": "3217293",
      "postDate": "06/04/2025 19:49:54",
      "content": "<p>I successfully created my first draft model for this competition, but I get some interesting behavior that I haven't seen before, nor do I understand what it means and how to solve it.</p>\n<p>While training my model (20 epochs), my training accuracy and loss are impeccable. They look beautiful in my graph below. However, my validation accuracy and loss make me want to vomit.</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F16391251%2Fe5b091e093bc4a959b4ef8f97347059b%2Fbad_validation.png?generation=1749066442410405&amp;alt=media\" alt=\"\"></p>\n<p>I compiled my model with the following line:</p>\n<pre><code>model.compile(=, =, metrics=[])\n</code></pre>\n<p>What are some potential causes for this behavior? What did I overlook to cause such a hot steamy pile of garbage?</p>",
      "rawMarkdown": "I successfully created my first draft model for this competition, but I get some interesting behavior that I haven't seen before, nor do I understand what it means and how to solve it.\n\nWhile training my model (20 epochs), my training accuracy and loss are impeccable. They look beautiful in my graph below. However, my validation accuracy and loss make me want to vomit.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F16391251%2Fe5b091e093bc4a959b4ef8f97347059b%2Fbad_validation.png?generation=1749066442410405&alt=media)\n\nI compiled my model with the following line:\n```\nmodel.compile(optimizer='adam', loss=\"binary_crossentropy\", metrics=['accuracy'])\n```\n\nWhat are some potential causes for this behavior? What did I overlook to cause such a hot steamy pile of garbage?",
      "votes": null
    },
    {
      "id": "3217307",
      "postDate": "06/04/2025 20:35:52",
      "content": "<p>em, looks like overfiting to me……<br>\ntry to strengthen the regularization, change learning rate, add normalization layers, so on…</p>",
      "rawMarkdown": "em, looks like overfiting to me......\ntry to strengthen the regularization, change learning rate, add normalization layers, so on...",
      "votes": null
    },
    {
      "id": "3217331",
      "postDate": "06/04/2025 22:19:15",
      "content": "<p>definetely overfitting<br>\nyour model is just memorizing the training data, learning patters that don't actually exist</p>",
      "rawMarkdown": "definetely overfitting\nyour model is just memorizing the training data, learning patters that don't actually exist",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3217307,
      "author_name": "eikyou",
      "author_url": "",
      "post_date": "06/04/2025 20:35:52",
      "content": "<p>em, looks like overfiting to me……<br>\ntry to strengthen the regularization, change learning rate, add normalization layers, so on…</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3217331,
      "author_name": "defdet",
      "author_url": "",
      "post_date": "06/04/2025 22:19:15",
      "content": "<p>definetely overfitting<br>\nyour model is just memorizing the training data, learning patters that don't actually exist</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3217293": "I successfully created my first draft model for this competition, but I get some interesting behavior that I haven't seen before, nor do I understand what it means and how to solve it.\n\nWhile training my model (20 epochs), my training accuracy and loss are impeccable. They look beautiful in my graph below. However, my validation accuracy and loss make me want to vomit.\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F16391251%2Fe5b091e093bc4a959b4ef8f97347059b%2Fbad_validation.png?generation=1749066442410405&alt=media)\n\nI compiled my model with the following line:\n```\nmodel.compile(optimizer='adam', loss=\"binary_crossentropy\", metrics=['accuracy'])\n```\n\nWhat are some potential causes for this behavior? What did I overlook to cause such a hot steamy pile of garbage?",
    "3217307": "em, looks like overfiting to me......\ntry to strengthen the regularization, change learning rate, add normalization layers, so on...",
    "3217331": "definetely overfitting\nyour model is just memorizing the training data, learning patters that don't actually exist"
  },
  "source": "meta"
}