{
  "id": 18776,
  "title": "Java",
  "url": "/competitions/second-annual-data-science-bowl/discussion/18776",
  "author_name": "knorra",
  "post_date": "2016-02-05T13:39:37.937000",
  "votes": 1,
  "comment_count": 5,
  "views": 977,
  "content": "<p>Is anyone using Java and would like to share experiences?</p>",
  "messages": [
    {
      "id": 107851,
      "postDate": "2016-02-13T03:39:58.810Z",
      "content": "<p>There are a few powerful tools in java, among them Spark and the Hadoop ecosystem. There is DeepLearning4J over at <a href=\"http://deeplearning4j.org/\">http://deeplearning4j.org/</a> which is powerful, flexible, and fast at handling all of your neural nets. </p>\n\n<p>I've tended to find that all of the code - in every language - is based on the linear algebra subsystems in BLAS (for CPU) or CUDA (for GPU) and provide abstraction layers into your code of choice. At some point, it comes down to which flavor you prefer.</p>",
      "rawMarkdown": "There are a few powerful tools in java, among them Spark and the Hadoop ecosystem. There is DeepLearning4J over at http://deeplearning4j.org/ which is powerful, flexible, and fast at handling all of your neural nets. \r\n\r\nI've tended to find that all of the code - in every language - is based on the linear algebra subsystems in BLAS (for CPU) or CUDA (for GPU) and provide abstraction layers into your code of choice. At some point, it comes down to which flavor you prefer.",
      "votes": 1
    },
    {
      "id": 106971,
      "postDate": "2016-02-05T13:39:37.937Z",
      "content": "<p>Is anyone using Java and would like to share experiences?</p>",
      "rawMarkdown": "Is anyone using Java and would like to share experiences?",
      "votes": 1
    },
    {
      "id": 108020,
      "postDate": "2016-02-14T21:49:01.593Z",
      "content": "<p>Good look. Writing everything again is very error prone. I know java and C++ but i dont use if often for ML. It doens speed up thinkgs since the core libraries are wirten in c++.</p>\n\n<p>[quote=TasDevil;108016]</p>\n\n<p>I'm using Java.  It took me almost two weeks to write a custom DICOM loader and get my head around the DICOM standard.  In retrospect it would have been easier to use Matlab or a library but I have learned exactly how the DICOM files work and if there were any issues with tag data type interpretations for example I could easily address that.   </p>\n\n<p>In my opinion using well written Java or C/C++ code would generally yield big speed gains which for me outweigh the extra work required to get things going especially for very large data sets.  </p>\n\n<p>[/quote]</p>",
      "rawMarkdown": "Good look. Writing everything again is very error prone. I know java and C++ but i dont use if often for ML. It doens speed up thinkgs since the core libraries are wirten in c++.\r\n\r\n\r\n[quote=TasDevil;108016]\r\n\r\nI'm using Java.  It took me almost two weeks to write a custom DICOM loader and get my head around the DICOM standard.  In retrospect it would have been easier to use Matlab or a library but I have learned exactly how the DICOM files work and if there were any issues with tag data type interpretations for example I could easily address that.   \r\n\r\nIn my opinion using well written Java or C/C++ code would generally yield big speed gains which for me outweigh the extra work required to get things going especially for very large data sets.  \r\n\r\n[/quote]\r\n"
    },
    {
      "id": 108016,
      "postDate": "2016-02-14T21:29:15.993Z",
      "content": "<p>I'm using Java.  It took me almost two weeks to write a custom DICOM loader and get my head around the DICOM standard.  In retrospect it would have been easier to use Matlab or a library but I have learned exactly how the DICOM files work and if there were any issues with tag data type interpretations for example I could easily address that.   </p>\n\n<p>In my opinion using well written Java or C/C++ code would generally yield big speed gains which for me outweigh the extra work required to get things going especially for very large data sets.  </p>",
      "rawMarkdown": "I'm using Java.  It took me almost two weeks to write a custom DICOM loader and get my head around the DICOM standard.  In retrospect it would have been easier to use Matlab or a library but I have learned exactly how the DICOM files work and if there were any issues with tag data type interpretations for example I could easily address that.   \r\n\r\nIn my opinion using well written Java or C/C++ code would generally yield big speed gains which for me outweigh the extra work required to get things going especially for very large data sets.  "
    },
    {
      "id": 107949,
      "postDate": "2016-02-14T03:02:21.387Z",
      "content": "<p>For DICOM image handling, fiji (imagej) is a good choice.  </p>",
      "rawMarkdown": "For DICOM image handling, fiji (imagej) is a good choice.  \r\n"
    },
    {
      "id": 107063,
      "postDate": "2016-02-06T11:34:54.010Z",
      "rawMarkdown": "",
      "votes": 1,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 107851,
      "author_name": "T-Rymz",
      "author_url": "",
      "post_date": "2016-02-13T03:39:58.810000",
      "content": "<p>There are a few powerful tools in java, among them Spark and the Hadoop ecosystem. There is DeepLearning4J over at <a href=\"http://deeplearning4j.org/\">http://deeplearning4j.org/</a> which is powerful, flexible, and fast at handling all of your neural nets. </p>\n\n<p>I've tended to find that all of the code - in every language - is based on the linear algebra subsystems in BLAS (for CPU) or CUDA (for GPU) and provide abstraction layers into your code of choice. At some point, it comes down to which flavor you prefer.</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 108020,
      "author_name": "Leustagos",
      "author_url": "",
      "post_date": "2016-02-14T21:49:01.593000",
      "content": "<p>Good look. Writing everything again is very error prone. I know java and C++ but i dont use if often for ML. It doens speed up thinkgs since the core libraries are wirten in c++.</p>\n\n<p>[quote=TasDevil;108016]</p>\n\n<p>I'm using Java.  It took me almost two weeks to write a custom DICOM loader and get my head around the DICOM standard.  In retrospect it would have been easier to use Matlab or a library but I have learned exactly how the DICOM files work and if there were any issues with tag data type interpretations for example I could easily address that.   </p>\n\n<p>In my opinion using well written Java or C/C++ code would generally yield big speed gains which for me outweigh the extra work required to get things going especially for very large data sets.  </p>\n\n<p>[/quote]</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 108016,
      "author_name": "TasDevil",
      "author_url": "",
      "post_date": "2016-02-14T21:29:15.993000",
      "content": "<p>I'm using Java.  It took me almost two weeks to write a custom DICOM loader and get my head around the DICOM standard.  In retrospect it would have been easier to use Matlab or a library but I have learned exactly how the DICOM files work and if there were any issues with tag data type interpretations for example I could easily address that.   </p>\n\n<p>In my opinion using well written Java or C/C++ code would generally yield big speed gains which for me outweigh the extra work required to get things going especially for very large data sets.  </p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 107949,
      "author_name": "IAsIam",
      "author_url": "",
      "post_date": "2016-02-14T03:02:21.387000",
      "content": "<p>For DICOM image handling, fiji (imagej) is a good choice.  </p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 107063,
      "author_name": "",
      "author_url": "",
      "post_date": "2016-02-06T11:34:54.010000",
      "content": "",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "107851": "There are a few powerful tools in java, among them Spark and the Hadoop ecosystem. There is DeepLearning4J over at http://deeplearning4j.org/ which is powerful, flexible, and fast at handling all of your neural nets. \r\n\r\nI've tended to find that all of the code - in every language - is based on the linear algebra subsystems in BLAS (for CPU) or CUDA (for GPU) and provide abstraction layers into your code of choice. At some point, it comes down to which flavor you prefer.",
    "106971": "Is anyone using Java and would like to share experiences?",
    "108020": "Good look. Writing everything again is very error prone. I know java and C++ but i dont use if often for ML. It doens speed up thinkgs since the core libraries are wirten in c++.\r\n\r\n\r\n[quote=TasDevil;108016]\r\n\r\nI'm using Java.  It took me almost two weeks to write a custom DICOM loader and get my head around the DICOM standard.  In retrospect it would have been easier to use Matlab or a library but I have learned exactly how the DICOM files work and if there were any issues with tag data type interpretations for example I could easily address that.   \r\n\r\nIn my opinion using well written Java or C/C++ code would generally yield big speed gains which for me outweigh the extra work required to get things going especially for very large data sets.  \r\n\r\n[/quote]\r\n",
    "108016": "I'm using Java.  It took me almost two weeks to write a custom DICOM loader and get my head around the DICOM standard.  In retrospect it would have been easier to use Matlab or a library but I have learned exactly how the DICOM files work and if there were any issues with tag data type interpretations for example I could easily address that.   \r\n\r\nIn my opinion using well written Java or C/C++ code would generally yield big speed gains which for me outweigh the extra work required to get things going especially for very large data sets.  ",
    "107949": "For DICOM image handling, fiji (imagej) is a good choice.  \r\n",
    "107063": ""
  }
}