{
  "id": 15060,
  "title": "More than one action per time in events?",
  "url": "/competitions/grasp-and-lift-eeg-detection/discussion/15060",
  "author_name": "",
  "post_date": "2015-07-05T22:19:08.630Z",
  "votes": null,
  "comment_count": 3,
  "views": 531,
  "content": "<p>Hello,</p>\n\n<p>This is my first competition, so i apologize if I am missing something.</p>\n<p>The data page says the data was sampled at 500 Hz, so is it correct to assume that each line on the data and events files is a measurement from the sensors at 1/500th of a second (or so) ?</p>\n\n<p>In the events file,subj1_series1_events, there are lines where there is a '1' in more than one column. Does this mean that both those are events are happening at that time ? I guess I started with the assumption that each time would have a single event associated with it.&nbsp;</p>\n<p>When we submit a file, can it have a '1' in more than one column?</p>\n<p>Chris s</p>",
  "messages": [
    {
      "id": "83488",
      "postDate": "07/05/2015 22:19:08",
      "content": "<p>Hello,</p>\n\n<p>This is my first competition, so i apologize if I am missing something.</p>\n<p>The data page says the data was sampled at 500 Hz, so is it correct to assume that each line on the data and events files is a measurement from the sensors at 1/500th of a second (or so) ?</p>\n\n<p>In the events file,subj1_series1_events, there are lines where there is a '1' in more than one column. Does this mean that both those are events are happening at that time ? I guess I started with the assumption that each time would have a single event associated with it.&nbsp;</p>\n<p>When we submit a file, can it have a '1' in more than one column?</p>\n<p>Chris s</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "83489",
      "postDate": "07/05/2015 23:10:29",
      "content": "<p>Chris,</p>\n<p>You are correct the time between samples is about 2ms.</p>\n<p>The events do overlap and you can make a submission with ones in multiple columns for the same row. For evaluation purposes each type of event is treated as a separate binary classification task. Your score is just the average of your score for each task.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "83491",
      "postDate": "07/05/2015 23:26:05",
      "content": "<p>So we could really be looking at 6 different models, one for each event.&nbsp;</p>\n<p>However, the events must take place is a given order, so only adjacent events can occur together.&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "83492",
      "postDate": "07/05/2015 23:50:28",
      "content": "<p>One model per event is one approach. It is probably the easiest to set up using standard supervised learning techniques. For the events occurring later in the trial it might be helpful to incorporate information about the other events.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 83489,
      "author_name": "devinanzelmo",
      "author_url": "",
      "post_date": "07/05/2015 23:10:29",
      "content": "<p>Chris,</p>\n<p>You are correct the time between samples is about 2ms.</p>\n<p>The events do overlap and you can make a submission with ones in multiple columns for the same row. For evaluation purposes each type of event is treated as a separate binary classification task. Your score is just the average of your score for each task.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 83491,
      "author_name": "chrisfs",
      "author_url": "",
      "post_date": "07/05/2015 23:26:05",
      "content": "<p>So we could really be looking at 6 different models, one for each event.&nbsp;</p>\n<p>However, the events must take place is a given order, so only adjacent events can occur together.&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 83492,
      "author_name": "devinanzelmo",
      "author_url": "",
      "post_date": "07/05/2015 23:50:28",
      "content": "<p>One model per event is one approach. It is probably the easiest to set up using standard supervised learning techniques. For the events occurring later in the trial it might be helpful to incorporate information about the other events.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "83488": "",
    "83489": "",
    "83491": "",
    "83492": ""
  },
  "source": "meta"
}