{
  "id": 15086,
  "title": "Improving the performance of CNN model",
  "url": "/competitions/diabetic-retinopathy-detection/discussion/15086",
  "author_name": "",
  "post_date": "2015-07-07T08:14:02.313Z",
  "votes": null,
  "comment_count": 1,
  "views": 1205,
  "content": "<p>I have resized the images to 256x256 and running the sample theano code given on the deep learning tutorial on the resized images; but I am getting not so good results. Any suggestions as in how many CNN layers, size of filters, learning rate, minibatch size etc. to get better results will be appreciated. It would be helpful if some  relevant links, articles can also be provided. Thanks in advance. \nP.S : I am new to deep learning and this is my first kaggle project and want to learn most from this project.</p>",
  "messages": [
    {
      "id": "83626",
      "postDate": "07/07/2015 08:14:02",
      "content": "<p>I have resized the images to 256x256 and running the sample theano code given on the deep learning tutorial on the resized images; but I am getting not so good results. Any suggestions as in how many CNN layers, size of filters, learning rate, minibatch size etc. to get better results will be appreciated. It would be helpful if some  relevant links, articles can also be provided. Thanks in advance. \nP.S : I am new to deep learning and this is my first kaggle project and want to learn most from this project.</p>",
      "rawMarkdown": "I have resized the images to 256x256 and running the sample theano code given on the deep learning tutorial on the resized images; but I am getting not so good results. Any suggestions as in how many CNN layers, size of filters, learning rate, minibatch size etc. to get better results will be appreciated. It would be helpful if some  relevant links, articles can also be provided. Thanks in advance. \r\nP.S : I am new to deep learning and this is my first kaggle project and want to learn most from this project.",
      "votes": null
    },
    {
      "id": "83750",
      "postDate": "07/08/2015 03:49:27",
      "content": "<p>This is a good reference for practical CNN advice: \n<a href=\"http://benanne.github.io/2015/03/17/plankton.html\">http://benanne.github.io/2015/03/17/plankton.html</a></p>",
      "rawMarkdown": "This is a good reference for practical CNN advice: \r\nhttp://benanne.github.io/2015/03/17/plankton.html",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 83750,
      "author_name": "dopelearner",
      "author_url": "",
      "post_date": "07/08/2015 03:49:27",
      "content": "<p>This is a good reference for practical CNN advice: \n<a href=\"http://benanne.github.io/2015/03/17/plankton.html\">http://benanne.github.io/2015/03/17/plankton.html</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "83626": "I have resized the images to 256x256 and running the sample theano code given on the deep learning tutorial on the resized images; but I am getting not so good results. Any suggestions as in how many CNN layers, size of filters, learning rate, minibatch size etc. to get better results will be appreciated. It would be helpful if some  relevant links, articles can also be provided. Thanks in advance. \r\nP.S : I am new to deep learning and this is my first kaggle project and want to learn most from this project.",
    "83750": "This is a good reference for practical CNN advice: \r\nhttp://benanne.github.io/2015/03/17/plankton.html"
  },
  "source": "meta"
}