{
  "id": 15924,
  "title": "Keras Index Batch Iterator",
  "url": "/competitions/grasp-and-lift-eeg-detection/discussion/15924",
  "author_name": "",
  "post_date": "2015-08-13T08:17:12.180Z",
  "votes": null,
  "comment_count": 1,
  "views": 1015,
  "content": "<p>I rebuild Tim Hochbergs (thanks for the script ;-) NN with Keras, but I could't figure out how to build a Index Batch Iterator in Keras. For now I'm loading all the data for a subject into the RAM, which works but is a pain. </p>\n\n<p>Has anyone done this, and can help me?\nThanks!</p>",
  "messages": [
    {
      "id": "89282",
      "postDate": "08/13/2015 08:17:12",
      "content": "<p>I rebuild Tim Hochbergs (thanks for the script ;-) NN with Keras, but I could't figure out how to build a Index Batch Iterator in Keras. For now I'm loading all the data for a subject into the RAM, which works but is a pain. </p>\n\n<p>Has anyone done this, and can help me?\nThanks!</p>",
      "rawMarkdown": "I rebuild Tim Hochbergs (thanks for the script ;-) NN with Keras, but I could't figure out how to build a Index Batch Iterator in Keras. For now I'm loading all the data for a subject into the RAM, which works but is a pain. \r\n\r\nHas anyone done this, and can help me?\r\nThanks!",
      "votes": null
    },
    {
      "id": "89307",
      "postDate": "08/13/2015 17:03:18",
      "content": "<p>For Keras, you can use model.train_on_batch(...) in order to avoid loading all data to the RAM. Of course, you will lose sklearn's APIs such as fit and predict</p>",
      "rawMarkdown": "For Keras, you can use model.train_on_batch(...) in order to avoid loading all data to the RAM. Of course, you will lose sklearn's APIs such as fit and predict",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 89307,
      "author_name": "songgc",
      "author_url": "",
      "post_date": "08/13/2015 17:03:18",
      "content": "<p>For Keras, you can use model.train_on_batch(...) in order to avoid loading all data to the RAM. Of course, you will lose sklearn's APIs such as fit and predict</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "89282": "I rebuild Tim Hochbergs (thanks for the script ;-) NN with Keras, but I could't figure out how to build a Index Batch Iterator in Keras. For now I'm loading all the data for a subject into the RAM, which works but is a pain. \r\n\r\nHas anyone done this, and can help me?\r\nThanks!",
    "89307": "For Keras, you can use model.train_on_batch(...) in order to avoid loading all data to the RAM. Of course, you will lose sklearn's APIs such as fit and predict"
  },
  "source": "meta"
}