{
  "id": 47622,
  "title": "Source Code",
  "url": "/competitions/favorita-grocery-sales-forecasting/discussion/47622",
  "author_name": "",
  "post_date": "2018-01-17T04:11:21.985210500Z",
  "votes": 43,
  "comment_count": 10,
  "views": 0,
  "content": "<p>I've released the source code I used for this model, you can view it here: <a href=\"https://github.com/sjvasquez/web-traffic-forecasting/tree/master/cf\">https://github.com/sjvasquez/web-traffic-forecasting/tree/master/cf</a></p>\n\n<p>These are the main differences (that I remember) from the code I used for the Web Traffic Forecasting competition:</p>\n\n<ul>\n<li>LSTM gating equations are used along the depth dimension</li>\n<li>Learning rate scheduling, batch size scheduling, momentum scheduling</li>\n<li>Fixed off-by-one bug in the encoder</li>\n<li>Fixed convolution padding bug in decoder</li>\n<li>Decoder augmented with a bidirectional LSTM</li>\n<li>Added feature engineering in the data processing pipeline</li>\n<li>Made some changes to batching and validation</li>\n<li>Output layer includes bernoulli parameter</li>\n</ul>\n\n<p>Should be easy to adapt the code for future forecasting competitions.</p>",
  "messages": [
    {
      "id": "269652",
      "postDate": "01/17/2018 04:11:21",
      "content": "<p>I've released the source code I used for this model, you can view it here: <a href=\"https://github.com/sjvasquez/web-traffic-forecasting/tree/master/cf\">https://github.com/sjvasquez/web-traffic-forecasting/tree/master/cf</a></p>\n\n<p>These are the main differences (that I remember) from the code I used for the Web Traffic Forecasting competition:</p>\n\n<ul>\n<li>LSTM gating equations are used along the depth dimension</li>\n<li>Learning rate scheduling, batch size scheduling, momentum scheduling</li>\n<li>Fixed off-by-one bug in the encoder</li>\n<li>Fixed convolution padding bug in decoder</li>\n<li>Decoder augmented with a bidirectional LSTM</li>\n<li>Added feature engineering in the data processing pipeline</li>\n<li>Made some changes to batching and validation</li>\n<li>Output layer includes bernoulli parameter</li>\n</ul>\n\n<p>Should be easy to adapt the code for future forecasting competitions.</p>",
      "rawMarkdown": "I've released the source code I used for this model, you can view it here: https://github.com/sjvasquez/web-traffic-forecasting/tree/master/cf\n\nThese are the main differences (that I remember) from the code I used for the Web Traffic Forecasting competition:\n\n - LSTM gating equations are used along the depth dimension\n - Learning rate scheduling, batch size scheduling, momentum scheduling\n - Fixed off-by-one bug in the encoder\n - Fixed convolution padding bug in decoder\n - Decoder augmented with a bidirectional LSTM\n - Added feature engineering in the data processing pipeline\n - Made some changes to batching and validation\n - Output layer includes bernoulli parameter\n\nShould be easy to adapt the code for future forecasting competitions.",
      "votes": null
    },
    {
      "id": "269659",
      "postDate": "01/17/2018 04:24:21",
      "content": "<p>Great work, thanks for sharing</p>",
      "rawMarkdown": "Great work, thanks for sharing",
      "votes": null
    },
    {
      "id": "269666",
      "postDate": "01/17/2018 04:32:58",
      "content": "<p>thanks for the sharing, sjv^_^</p>",
      "rawMarkdown": "thanks for the sharing, sjv^_^",
      "votes": null
    },
    {
      "id": "269674",
      "postDate": "01/17/2018 05:02:32",
      "content": "<p>This is impressive! Thanks for sharing, sjv!</p>",
      "rawMarkdown": "This is impressive! Thanks for sharing, sjv!",
      "votes": null
    },
    {
      "id": "269683",
      "postDate": "01/17/2018 05:20:33",
      "content": "<p>Really great work! hope you feel better now, thanks for sharing!</p>",
      "rawMarkdown": "Really great work! hope you feel better now, thanks for sharing!",
      "votes": null
    },
    {
      "id": "269738",
      "postDate": "01/17/2018 07:41:23",
      "content": "<p>Thanks for sharing, hope you feel better!</p>",
      "rawMarkdown": "Thanks for sharing, hope you feel better!",
      "votes": null
    },
    {
      "id": "269766",
      "postDate": "01/17/2018 08:55:10",
      "content": "<p>thank you very much, it's always great to see someone so generous</p>",
      "rawMarkdown": "thank you very much, it's always great to see someone so generous",
      "votes": null
    },
    {
      "id": "269895",
      "postDate": "01/17/2018 13:44:07",
      "content": "<p>Great work, thanks for sharing</p>",
      "rawMarkdown": "Great work, thanks for sharing",
      "votes": null
    },
    {
      "id": "269999",
      "postDate": "01/17/2018 16:30:17",
      "content": "<p>Amazing solution.  Thank you for sharing !!!</p>",
      "rawMarkdown": "Amazing solution.  Thank you for sharing !!!",
      "votes": null
    },
    {
      "id": "270003",
      "postDate": "01/17/2018 16:38:33",
      "content": "<p>You are so good!. Thanks very much. I think your sharing makes kaggle community more charming! I really hope you can occasionally participate some competition in the future when you got new ideas or new models. Just like the the author of xgboost. </p>",
      "rawMarkdown": "You are so good!. Thanks very much. I think your sharing makes kaggle community more charming! I really hope you can occasionally participate some competition in the future when you got new ideas or new models. Just like the the author of xgboost.",
      "votes": null
    },
    {
      "id": "270323",
      "postDate": "01/18/2018 04:53:00",
      "content": "<p>Thanks for sharing!</p>",
      "rawMarkdown": "Thanks for sharing!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 269659,
      "author_name": "shixw125",
      "author_url": "",
      "post_date": "01/17/2018 04:24:21",
      "content": "<p>Great work, thanks for sharing</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 269666,
      "author_name": "sushize",
      "author_url": "",
      "post_date": "01/17/2018 04:32:58",
      "content": "<p>thanks for the sharing, sjv^_^</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 269674,
      "author_name": "cczaixian",
      "author_url": "",
      "post_date": "01/17/2018 05:02:32",
      "content": "<p>This is impressive! Thanks for sharing, sjv!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 269683,
      "author_name": "gerryl",
      "author_url": "",
      "post_date": "01/17/2018 05:20:33",
      "content": "<p>Really great work! hope you feel better now, thanks for sharing!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 269738,
      "author_name": "cpmpml",
      "author_url": "",
      "post_date": "01/17/2018 07:41:23",
      "content": "<p>Thanks for sharing, hope you feel better!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 269766,
      "author_name": "steelrose",
      "author_url": "",
      "post_date": "01/17/2018 08:55:10",
      "content": "<p>thank you very much, it's always great to see someone so generous</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 269895,
      "author_name": "infzero",
      "author_url": "",
      "post_date": "01/17/2018 13:44:07",
      "content": "<p>Great work, thanks for sharing</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 269999,
      "author_name": "paulz11",
      "author_url": "",
      "post_date": "01/17/2018 16:30:17",
      "content": "<p>Amazing solution.  Thank you for sharing !!!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 270003,
      "author_name": "luciusyu",
      "author_url": "",
      "post_date": "01/17/2018 16:38:33",
      "content": "<p>You are so good!. Thanks very much. I think your sharing makes kaggle community more charming! I really hope you can occasionally participate some competition in the future when you got new ideas or new models. Just like the the author of xgboost. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 270323,
      "author_name": "beletecheneke",
      "author_url": "",
      "post_date": "01/18/2018 04:53:00",
      "content": "<p>Thanks for sharing!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "269652": "I've released the source code I used for this model, you can view it here: https://github.com/sjvasquez/web-traffic-forecasting/tree/master/cf\n\nThese are the main differences (that I remember) from the code I used for the Web Traffic Forecasting competition:\n\n - LSTM gating equations are used along the depth dimension\n - Learning rate scheduling, batch size scheduling, momentum scheduling\n - Fixed off-by-one bug in the encoder\n - Fixed convolution padding bug in decoder\n - Decoder augmented with a bidirectional LSTM\n - Added feature engineering in the data processing pipeline\n - Made some changes to batching and validation\n - Output layer includes bernoulli parameter\n\nShould be easy to adapt the code for future forecasting competitions.",
    "269659": "Great work, thanks for sharing",
    "269666": "thanks for the sharing, sjv^_^",
    "269674": "This is impressive! Thanks for sharing, sjv!",
    "269683": "Really great work! hope you feel better now, thanks for sharing!",
    "269738": "Thanks for sharing, hope you feel better!",
    "269766": "thank you very much, it's always great to see someone so generous",
    "269895": "Great work, thanks for sharing",
    "269999": "Amazing solution.  Thank you for sharing !!!",
    "270003": "You are so good!. Thanks very much. I think your sharing makes kaggle community more charming! I really hope you can occasionally participate some competition in the future when you got new ideas or new models. Just like the the author of xgboost.",
    "270323": "Thanks for sharing!"
  },
  "source": "meta"
}