{
  "id": 113121,
  "title": "Use of Automated Machine Learning Tools (AMLT)",
  "url": "/competitions/ashrae-energy-prediction/discussion/113121",
  "author_name": "",
  "post_date": "2019-10-17T05:39:09.862297500Z",
  "votes": 2,
  "comment_count": 7,
  "views": 0,
  "content": "<p>I want to ask about what's the difference between Automated Machine Learning Tools (AMLT) and conventional Machine Learning Tools.</p>\n\n<p>As I didn't find a definition of what AMLT is, I prefer to ask here in order to learn a bit more.</p>\n\n<p>I mean, in instance, the scikit-learn seems pretty automated for me. I consider it conventional, in the way I consider automated something you put the data into and without doing too much, you get a result.</p>\n\n<p>Then, what's the limit between automated and conventional?</p>",
  "messages": [
    {
      "id": "651152",
      "postDate": "10/17/2019 05:39:09",
      "content": "<p>I want to ask about what's the difference between Automated Machine Learning Tools (AMLT) and conventional Machine Learning Tools.</p>\n\n<p>As I didn't find a definition of what AMLT is, I prefer to ask here in order to learn a bit more.</p>\n\n<p>I mean, in instance, the scikit-learn seems pretty automated for me. I consider it conventional, in the way I consider automated something you put the data into and without doing too much, you get a result.</p>\n\n<p>Then, what's the limit between automated and conventional?</p>",
      "rawMarkdown": "I want to ask about what's the difference between Automated Machine Learning Tools (AMLT) and conventional Machine Learning Tools.\n\nAs I didn't find a definition of what AMLT is, I prefer to ask here in order to learn a bit more.\n\nI mean, in instance, the scikit-learn seems pretty automated for me. I consider it conventional, in the way I consider automated something you put the data into and without doing too much, you get a result.\n\nThen, what's the limit between automated and conventional?",
      "votes": null
    },
    {
      "id": "651159",
      "postDate": "10/17/2019 06:04:04",
      "content": "<p>Correct me if I am wrong, but Automated Machine Learning could mean a lot of things. \nAt the lowest, it could mean automatic algorithm selection/hyperparameter optimisation like autosklearn.\nIn the middle, it could mean automatic model architecture search as seen in autokeras.\nAt the highest, it could mean automatic configuration of model pipelines. as seen in TPOT </p>",
      "rawMarkdown": "Correct me if I am wrong, but Automated Machine Learning could mean a lot of things. \nAt the lowest, it could mean automatic algorithm selection/hyperparameter optimisation like autosklearn.\nIn the middle, it could mean automatic model architecture search as seen in autokeras.\nAt the highest, it could mean automatic configuration of model pipelines. as seen in TPOT",
      "votes": null
    },
    {
      "id": "652760",
      "postDate": "10/19/2019 11:27:28",
      "content": "<p>You may want to refer to the paper : <a href=\"https://arxiv.org/pdf/1810.13306.pdf\">Taking the Human out of Learning Applications: A Survey on Automated Machine Learning</a></p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F238237%2F066da0b95e860765de16f1331b8a3230%2F1810_13306_pdf.jpg?generation=1571484444020944&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "You may want to refer to the paper : [Taking the Human out of Learning Applications: A Survey on Automated Machine Learning](https://arxiv.org/pdf/1810.13306.pdf)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F238237%2F066da0b95e860765de16f1331b8a3230%2F1810_13306_pdf.jpg?generation=1571484444020944&amp;alt=media)",
      "votes": null
    },
    {
      "id": "652812",
      "postDate": "10/19/2019 12:58:42",
      "content": "<p>I do all my backprop by hand and expect everyone else to do the same.</p>",
      "rawMarkdown": "I do all my backprop by hand and expect everyone else to do the same.",
      "votes": null
    },
    {
      "id": "652914",
      "postDate": "10/19/2019 16:02:48",
      "content": "<p>If you had read the rules, there are two categories. The ones who do all 'by hand' and the ones who use  AMLT. The latter ones are not eligible to win any prizes.</p>\n\n<p>But all people in these two categories uses tools. Al least me, I only do some diagrams by hand. I use tools, and all tools automate things. What things are not permitted to automate, is the root of my question.</p>",
      "rawMarkdown": "If you had read the rules, there are two categories. The ones who do all 'by hand' and the ones who use  AMLT. The latter ones are not eligible to win any prizes.\n\nBut all people in these two categories uses tools. Al least me, I only do some diagrams by hand. I use tools, and all tools automate things. What things are not permitted to automate, is the root of my question.",
      "votes": null
    },
    {
      "id": "656085",
      "postDate": "10/23/2019 22:17:38",
      "content": "<p>It should be more clearly defined meaning of AMLT by organizers. </p>",
      "rawMarkdown": "It should be more clearly defined meaning of AMLT by organizers.",
      "votes": null
    },
    {
      "id": "656551",
      "postDate": "10/24/2019 11:33:40",
      "content": "<p>talking about automated tools, this may be useful\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F113660%2Fb42c068939fff981509648c8edd3d8e2%2FSelection_067.png?generation=1571916818262159&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "talking about automated tools, this may be useful\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F113660%2Fb42c068939fff981509648c8edd3d8e2%2FSelection_067.png?generation=1571916818262159&amp;alt=media)",
      "votes": null
    },
    {
      "id": "662791",
      "postDate": "11/01/2019 02:21:48",
      "content": "<p>What about using an AMLT to inform the kind of model to use in the final submission? Is that allowed?</p>",
      "rawMarkdown": "What about using an AMLT to inform the kind of model to use in the final submission? Is that allowed?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 651159,
      "author_name": "tiannnn",
      "author_url": "",
      "post_date": "10/17/2019 06:04:04",
      "content": "<p>Correct me if I am wrong, but Automated Machine Learning could mean a lot of things. \nAt the lowest, it could mean automatic algorithm selection/hyperparameter optimisation like autosklearn.\nIn the middle, it could mean automatic model architecture search as seen in autokeras.\nAt the highest, it could mean automatic configuration of model pipelines. as seen in TPOT </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 652760,
      "author_name": "arnabbiswas1",
      "author_url": "",
      "post_date": "10/19/2019 11:27:28",
      "content": "<p>You may want to refer to the paper : <a href=\"https://arxiv.org/pdf/1810.13306.pdf\">Taking the Human out of Learning Applications: A Survey on Automated Machine Learning</a></p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F238237%2F066da0b95e860765de16f1331b8a3230%2F1810_13306_pdf.jpg?generation=1571484444020944&amp;alt=media\" alt=\"\"></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 652812,
      "author_name": "sdoria",
      "author_url": "",
      "post_date": "10/19/2019 12:58:42",
      "content": "<p>I do all my backprop by hand and expect everyone else to do the same.</p>",
      "votes": null,
      "replies": [
        {
          "id": 652914,
          "author_name": "juanmah",
          "author_url": "",
          "post_date": "10/19/2019 16:02:48",
          "content": "<p>If you had read the rules, there are two categories. The ones who do all 'by hand' and the ones who use  AMLT. The latter ones are not eligible to win any prizes.</p>\n\n<p>But all people in these two categories uses tools. Al least me, I only do some diagrams by hand. I use tools, and all tools automate things. What things are not permitted to automate, is the root of my question.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 656085,
      "author_name": "gerlav",
      "author_url": "",
      "post_date": "10/23/2019 22:17:38",
      "content": "<p>It should be more clearly defined meaning of AMLT by organizers. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 656551,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "10/24/2019 11:33:40",
      "content": "<p>talking about automated tools, this may be useful\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F113660%2Fb42c068939fff981509648c8edd3d8e2%2FSelection_067.png?generation=1571916818262159&amp;alt=media\" alt=\"\"></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 662791,
      "author_name": "marcwigzell",
      "author_url": "",
      "post_date": "11/01/2019 02:21:48",
      "content": "<p>What about using an AMLT to inform the kind of model to use in the final submission? Is that allowed?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "651152": "I want to ask about what's the difference between Automated Machine Learning Tools (AMLT) and conventional Machine Learning Tools.\n\nAs I didn't find a definition of what AMLT is, I prefer to ask here in order to learn a bit more.\n\nI mean, in instance, the scikit-learn seems pretty automated for me. I consider it conventional, in the way I consider automated something you put the data into and without doing too much, you get a result.\n\nThen, what's the limit between automated and conventional?",
    "651159": "Correct me if I am wrong, but Automated Machine Learning could mean a lot of things. \nAt the lowest, it could mean automatic algorithm selection/hyperparameter optimisation like autosklearn.\nIn the middle, it could mean automatic model architecture search as seen in autokeras.\nAt the highest, it could mean automatic configuration of model pipelines. as seen in TPOT",
    "652760": "You may want to refer to the paper : [Taking the Human out of Learning Applications: A Survey on Automated Machine Learning](https://arxiv.org/pdf/1810.13306.pdf)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F238237%2F066da0b95e860765de16f1331b8a3230%2F1810_13306_pdf.jpg?generation=1571484444020944&amp;alt=media)",
    "652812": "I do all my backprop by hand and expect everyone else to do the same.",
    "652914": "If you had read the rules, there are two categories. The ones who do all 'by hand' and the ones who use  AMLT. The latter ones are not eligible to win any prizes.\n\nBut all people in these two categories uses tools. Al least me, I only do some diagrams by hand. I use tools, and all tools automate things. What things are not permitted to automate, is the root of my question.",
    "656085": "It should be more clearly defined meaning of AMLT by organizers.",
    "656551": "talking about automated tools, this may be useful\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F113660%2Fb42c068939fff981509648c8edd3d8e2%2FSelection_067.png?generation=1571916818262159&amp;alt=media)",
    "662791": "What about using an AMLT to inform the kind of model to use in the final submission? Is that allowed?"
  },
  "source": "meta"
}