{
  "id": 471487,
  "title": "CNN - Some advice",
  "url": "/competitions/UBC-OCEAN/discussion/471487",
  "author_name": "Stanislas TOUZET",
  "post_date": "2024-01-28T14:19:39.715000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi everyone !</p>\n<p>I have this competition as a project for my studies and I'm trying to build a CNN model on the dataset and have some questions :</p>\n<ul>\n<li>I don't really get how I can choose my activation function between conv layers : is there some intuition to have to choose it or is it just by testing different functions, so by hyperparameter optimization ?</li>\n<li>Is conv3d possible to use ? I've seen that it is used for videos and sequence of images but I also heard it is also used for medical images but I don't get in this context if conv3d is relevant ?</li>\n<li>Is there a kind of procedure for hyperparameter optimization, I mean, is there a good way to do it in the sense I should start with some parameters first or I can do it in the order I want ?</li>\n<li>Also, should I normalize the images or rescale them ?</li>\n</ul>\n<p>Sorry if my questions sound silly, just starting my machine learning journey :)</p>\n<p>Bless you</p>",
  "messages": [
    {
      "id": 2623994,
      "postDate": "2024-01-28T14:19:39.717Z",
      "content": "<p>Hi everyone !</p>\n<p>I have this competition as a project for my studies and I'm trying to build a CNN model on the dataset and have some questions :</p>\n<ul>\n<li>I don't really get how I can choose my activation function between conv layers : is there some intuition to have to choose it or is it just by testing different functions, so by hyperparameter optimization ?</li>\n<li>Is conv3d possible to use ? I've seen that it is used for videos and sequence of images but I also heard it is also used for medical images but I don't get in this context if conv3d is relevant ?</li>\n<li>Is there a kind of procedure for hyperparameter optimization, I mean, is there a good way to do it in the sense I should start with some parameters first or I can do it in the order I want ?</li>\n<li>Also, should I normalize the images or rescale them ?</li>\n</ul>\n<p>Sorry if my questions sound silly, just starting my machine learning journey :)</p>\n<p>Bless you</p>",
      "rawMarkdown": "Hi everyone !\n\nI have this competition as a project for my studies and I'm trying to build a CNN model on the dataset and have some questions :\n- I don't really get how I can choose my activation function between conv layers : is there some intuition to have to choose it or is it just by testing different functions, so by hyperparameter optimization ?\n- Is conv3d possible to use ? I've seen that it is used for videos and sequence of images but I also heard it is also used for medical images but I don't get in this context if conv3d is relevant ?\n- Is there a kind of procedure for hyperparameter optimization, I mean, is there a good way to do it in the sense I should start with some parameters first or I can do it in the order I want ?\n- Also, should I normalize the images or rescale them ?\n\nSorry if my questions sound silly, just starting my machine learning journey :)\n\nBless you"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2623994": "Hi everyone !\n\nI have this competition as a project for my studies and I'm trying to build a CNN model on the dataset and have some questions :\n- I don't really get how I can choose my activation function between conv layers : is there some intuition to have to choose it or is it just by testing different functions, so by hyperparameter optimization ?\n- Is conv3d possible to use ? I've seen that it is used for videos and sequence of images but I also heard it is also used for medical images but I don't get in this context if conv3d is relevant ?\n- Is there a kind of procedure for hyperparameter optimization, I mean, is there a good way to do it in the sense I should start with some parameters first or I can do it in the order I want ?\n- Also, should I normalize the images or rescale them ?\n\nSorry if my questions sound silly, just starting my machine learning journey :)\n\nBless you"
  }
}