{
  "id": 93162,
  "title": "WTTE-RNN",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/93162",
  "author_name": "delai50",
  "post_date": "2019-05-23T21:09:35.758000",
  "votes": 5,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Hi guys,</p>\n\n<p>Is anybody using this type of network? I heard Paweł Jankiewicz talking about it in a Kaggle Days Meetup, and it seems a very good approach. However, I am not well experienced with NN and I still don't get the point. Hope that somebody can create and share a good kernel with that!</p>\n\n<p>Here some links:\n<a href=\"https://github.com/ragulpr/wtte-rnn\">https://github.com/ragulpr/wtte-rnn</a>\n<a href=\"https://github.com/gm-spacagna/deep-ttf\">https://github.com/gm-spacagna/deep-ttf</a></p>\n\n<p>Regards!</p>",
  "messages": [
    {
      "id": 536034,
      "postDate": "2019-05-23T21:09:35.760Z",
      "content": "<p>Hi guys,</p>\n\n<p>Is anybody using this type of network? I heard Paweł Jankiewicz talking about it in a Kaggle Days Meetup, and it seems a very good approach. However, I am not well experienced with NN and I still don't get the point. Hope that somebody can create and share a good kernel with that!</p>\n\n<p>Here some links:\n<a href=\"https://github.com/ragulpr/wtte-rnn\">https://github.com/ragulpr/wtte-rnn</a>\n<a href=\"https://github.com/gm-spacagna/deep-ttf\">https://github.com/gm-spacagna/deep-ttf</a></p>\n\n<p>Regards!</p>",
      "rawMarkdown": "Hi guys,\n\nIs anybody using this type of network? I heard Paweł Jankiewicz talking about it in a Kaggle Days Meetup, and it seems a very good approach. However, I am not well experienced with NN and I still don't get the point. Hope that somebody can create and share a good kernel with that!\n\nHere some links:\nhttps://github.com/ragulpr/wtte-rnn\nhttps://github.com/gm-spacagna/deep-ttf\n\nRegards!",
      "votes": 5
    },
    {
      "id": 536077,
      "postDate": "2019-05-23T22:48:36.053Z",
      "content": "<p>It's a nice explanation!</p>\n\n<p>But it looks like you need a contiguous train of events that let you construct a distribution for the next event. Unfortunately, the LB data sample random rtf's. I could be interpreting wrongly, though.</p>",
      "rawMarkdown": "It's a nice explanation!\n\nBut it looks like you need a contiguous train of events that let you construct a distribution for the next event. Unfortunately, the LB data sample random rtf's. I could be interpreting wrongly, though.",
      "votes": 1
    },
    {
      "id": 536612,
      "postDate": "2019-05-24T20:13:46.607Z",
      "content": "<p>Happy to team up to explore this approach further if anyone interested in</p>",
      "rawMarkdown": "Happy to team up to explore this approach further if anyone interested in",
      "votes": 2,
      "replies": [
        {
          "id": 536699,
          "postDate": "2019-05-25T03:22:38.317Z",
          "content": "<p><a href=\"/fernandoramacciotti\">@fernandoramacciotti</a>  and <a href=\"/smerrllo\">@smerrllo</a>  I would love to team up to explore WTTE-RNN</p>",
          "rawMarkdown": "@fernandoramacciotti  and @smerrllo  I would love to team up to explore WTTE-RNN",
          "votes": 1
        },
        {
          "id": 536837,
          "postDate": "2019-05-25T11:41:08.010Z",
          "content": "<p>That would be awesome</p>",
          "rawMarkdown": "That would be awesome"
        }
      ]
    },
    {
      "id": 536096,
      "postDate": "2019-05-24T00:36:52.060Z",
      "content": "<p>I tried it and seemed promising, but to be honest didn't spend much time tuning it</p>\n\n<p>Please see my attempt below\n<a href=\"https://www.kaggle.com/fernandoramacciotti/weibull-time-to-failure\">https://www.kaggle.com/fernandoramacciotti/weibull-time-to-failure</a></p>",
      "rawMarkdown": "I tried it and seemed promising, but to be honest didn't spend much time tuning it\n\nPlease see my attempt below\nhttps://www.kaggle.com/fernandoramacciotti/weibull-time-to-failure\n",
      "votes": 2
    },
    {
      "id": 536284,
      "postDate": "2019-05-24T07:26:05.807Z",
      "rawMarkdown": "",
      "votes": 2,
      "isDeleted": true,
      "replies": [
        {
          "id": 536694,
          "postDate": "2019-05-25T02:58:53.220Z",
          "content": "<p><a href=\"/smerrllo\">@smerrllo</a> I definitely would like to see how you applied this to medical data.  I am very interested and would love get some of your input related to that project. </p>",
          "rawMarkdown": "@smerrllo I definitely would like to see how you applied this to medical data.  I am very interested and would love get some of your input related to that project. "
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 536077,
      "author_name": "pete",
      "author_url": "",
      "post_date": "2019-05-23T22:48:36.053000",
      "content": "<p>It's a nice explanation!</p>\n\n<p>But it looks like you need a contiguous train of events that let you construct a distribution for the next event. Unfortunately, the LB data sample random rtf's. I could be interpreting wrongly, though.</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 536612,
      "author_name": "Fernando Ramacciotti",
      "author_url": "",
      "post_date": "2019-05-24T20:13:46.607000",
      "content": "<p>Happy to team up to explore this approach further if anyone interested in</p>",
      "votes": 2,
      "replies": [
        {
          "id": 536699,
          "author_name": "William Green",
          "author_url": "",
          "post_date": "2019-05-25T03:22:38.317000",
          "content": "<p><a href=\"/fernandoramacciotti\">@fernandoramacciotti</a>  and <a href=\"/smerrllo\">@smerrllo</a>  I would love to team up to explore WTTE-RNN</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 536837,
          "author_name": "Fernando Ramacciotti",
          "author_url": "",
          "post_date": "2019-05-25T11:41:08.010000",
          "content": "<p>That would be awesome</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 536096,
      "author_name": "Fernando Ramacciotti",
      "author_url": "",
      "post_date": "2019-05-24T00:36:52.060000",
      "content": "<p>I tried it and seemed promising, but to be honest didn't spend much time tuning it</p>\n\n<p>Please see my attempt below\n<a href=\"https://www.kaggle.com/fernandoramacciotti/weibull-time-to-failure\">https://www.kaggle.com/fernandoramacciotti/weibull-time-to-failure</a></p>",
      "votes": 2,
      "replies": []
    },
    {
      "id": 536284,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-05-24T07:26:05.807000",
      "content": "",
      "votes": 2,
      "replies": [
        {
          "id": 536694,
          "author_name": "William Green",
          "author_url": "",
          "post_date": "2019-05-25T02:58:53.220000",
          "content": "<p><a href=\"/smerrllo\">@smerrllo</a> I definitely would like to see how you applied this to medical data.  I am very interested and would love get some of your input related to that project. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "536034": "Hi guys,\n\nIs anybody using this type of network? I heard Paweł Jankiewicz talking about it in a Kaggle Days Meetup, and it seems a very good approach. However, I am not well experienced with NN and I still don't get the point. Hope that somebody can create and share a good kernel with that!\n\nHere some links:\nhttps://github.com/ragulpr/wtte-rnn\nhttps://github.com/gm-spacagna/deep-ttf\n\nRegards!",
    "536077": "It's a nice explanation!\n\nBut it looks like you need a contiguous train of events that let you construct a distribution for the next event. Unfortunately, the LB data sample random rtf's. I could be interpreting wrongly, though.",
    "536612": "Happy to team up to explore this approach further if anyone interested in",
    "536096": "I tried it and seemed promising, but to be honest didn't spend much time tuning it\n\nPlease see my attempt below\nhttps://www.kaggle.com/fernandoramacciotti/weibull-time-to-failure\n",
    "536284": ""
  }
}