{
  "id": 74149,
  "title": "Temporal Convolutions Network",
  "url": "/competitions/quora-insincere-questions-classification/discussion/74149",
  "author_name": "",
  "post_date": "2018-12-09T08:29:39.947193800Z",
  "votes": 14,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Anyone tried Temporal Convolutions Network (TCN) yet?</p>\n\n<p>I illustrated in the kernels section:</p>\n\n<p><a href=\"https://www.kaggle.com/christofhenkel/temporal-convolutional-network\">TCN with Keras</a></p>\n\n<p>From literature TCN seems very powerful yet time efficient.  But might need more tuning than I did to perform good. You can stack the TCN layers or even combine with LSTM or GRU. Would be happy to see someone posts some results here. </p>",
  "messages": [
    {
      "id": "435990",
      "postDate": "12/09/2018 08:29:39",
      "content": "<p>Anyone tried Temporal Convolutions Network (TCN) yet?</p>\n\n<p>I illustrated in the kernels section:</p>\n\n<p><a href=\"https://www.kaggle.com/christofhenkel/temporal-convolutional-network\">TCN with Keras</a></p>\n\n<p>From literature TCN seems very powerful yet time efficient.  But might need more tuning than I did to perform good. You can stack the TCN layers or even combine with LSTM or GRU. Would be happy to see someone posts some results here. </p>",
      "rawMarkdown": "Anyone tried Temporal Convolutions Network (TCN) yet?\n\nI illustrated in the kernels section:\n\n[TCN with Keras][1]\n\nFrom literature TCN seems very powerful yet time efficient.  But might need more tuning than I did to perform good. You can stack the TCN layers or even combine with LSTM or GRU. Would be happy to see someone posts some results here. \n\n  [1]: https://www.kaggle.com/christofhenkel/temporal-convolutional-network",
      "votes": null
    },
    {
      "id": "436031",
      "postDate": "12/09/2018 11:19:22",
      "content": "<p>Thanks Dieter for the post. </p>\n\n<p>I tried to stack TCN layers earlier in the competition, but didn't get very good results.   As you say, It might require some tuning (I didn't try to combine it with LSTM tough) </p>",
      "rawMarkdown": "Thanks Dieter for the post. \n\nI tried to stack TCN layers earlier in the competition, but didn't get very good results.   As you say, It might require some tuning (I didn't try to combine it with LSTM tough)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 436031,
      "author_name": "serigne",
      "author_url": "",
      "post_date": "12/09/2018 11:19:22",
      "content": "<p>Thanks Dieter for the post. </p>\n\n<p>I tried to stack TCN layers earlier in the competition, but didn't get very good results.   As you say, It might require some tuning (I didn't try to combine it with LSTM tough) </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "435990": "Anyone tried Temporal Convolutions Network (TCN) yet?\n\nI illustrated in the kernels section:\n\n[TCN with Keras][1]\n\nFrom literature TCN seems very powerful yet time efficient.  But might need more tuning than I did to perform good. You can stack the TCN layers or even combine with LSTM or GRU. Would be happy to see someone posts some results here. \n\n  [1]: https://www.kaggle.com/christofhenkel/temporal-convolutional-network",
    "436031": "Thanks Dieter for the post. \n\nI tried to stack TCN layers earlier in the competition, but didn't get very good results.   As you say, It might require some tuning (I didn't try to combine it with LSTM tough)"
  },
  "source": "meta"
}