{
  "id": 79126,
  "title": "Seriously attention working guys?",
  "url": "/competitions/quora-insincere-questions-classification/discussion/79126",
  "author_name": "aintnosunshine",
  "post_date": "2019-01-31T07:23:30.887000",
  "votes": 3,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I have tried attention, with and without masking. Every time I use that, results are not up to, those without attention. I tried on some other dataset, same performance issue was there. So, is it really worth using attention? Those who are familiar with attention, any idea suggestions.</p>",
  "messages": [
    {
      "id": 464090,
      "postDate": "2019-01-31T07:23:30.887Z",
      "content": "<p>I have tried attention, with and without masking. Every time I use that, results are not up to, those without attention. I tried on some other dataset, same performance issue was there. So, is it really worth using attention? Those who are familiar with attention, any idea suggestions.</p>",
      "rawMarkdown": "I have tried attention, with and without masking. Every time I use that, results are not up to, those without attention. I tried on some other dataset, same performance issue was there. So, is it really worth using attention? Those who are familiar with attention, any idea suggestions.",
      "votes": 3
    },
    {
      "id": 464391,
      "postDate": "2019-01-31T19:01:33.523Z",
      "content": "<p>So what are you using instead of attention? Pooling? Last output?</p>",
      "rawMarkdown": "So what are you using instead of attention? Pooling? Last output?",
      "votes": 1,
      "replies": [
        {
          "id": 464499,
          "postDate": "2019-02-01T01:34:32.477Z",
          "content": "<p>I am not using any attention. I am using max pooling, that works better. </p>",
          "rawMarkdown": "I am not using any attention. I am using max pooling, that works better. ",
          "votes": 1
        }
      ]
    },
    {
      "id": 464141,
      "postDate": "2019-01-31T08:50:40.280Z",
      "content": "<p>I tried Attention (taken from public kernel) between recurrent layers (LSTM, GRU) and pooling layers:\nDid not harm, did not help, so it was neutral.\nI only did very few experiments though.</p>",
      "rawMarkdown": "I tried Attention (taken from public kernel) between recurrent layers (LSTM, GRU) and pooling layers:\nDid not harm, did not help, so it was neutral.\nI only did very few experiments though.",
      "votes": 5,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 464391,
      "author_name": "Max Schumacher",
      "author_url": "",
      "post_date": "2019-01-31T19:01:33.523000",
      "content": "<p>So what are you using instead of attention? Pooling? Last output?</p>",
      "votes": 1,
      "replies": [
        {
          "id": 464499,
          "author_name": "aintnosunshine",
          "author_url": "",
          "post_date": "2019-02-01T01:34:32.477000",
          "content": "<p>I am not using any attention. I am using max pooling, that works better. </p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 464141,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-01-31T08:50:40.280000",
      "content": "<p>I tried Attention (taken from public kernel) between recurrent layers (LSTM, GRU) and pooling layers:\nDid not harm, did not help, so it was neutral.\nI only did very few experiments though.</p>",
      "votes": 5,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "464090": "I have tried attention, with and without masking. Every time I use that, results are not up to, those without attention. I tried on some other dataset, same performance issue was there. So, is it really worth using attention? Those who are familiar with attention, any idea suggestions.",
    "464391": "So what are you using instead of attention? Pooling? Last output?",
    "464141": "I tried Attention (taken from public kernel) between recurrent layers (LSTM, GRU) and pooling layers:\nDid not harm, did not help, so it was neutral.\nI only did very few experiments though."
  }
}