{
  "id": 15053,
  "title": "Definition of a Frame ",
  "url": "/competitions/grasp-and-lift-eeg-detection/discussion/15053",
  "author_name": "",
  "post_date": "2015-07-05T01:37:37.247Z",
  "votes": null,
  "comment_count": 2,
  "views": 605,
  "content": "<p>Hello!</p>\n<p>In this project we're collecting data at 500Hz. &nbsp;Each sample is 2ms. &nbsp;As I &nbsp;understand it, the&nbsp;detection window in the discussion was +/- 75 samples wide. &nbsp;Does this mean that I can run with a lag of 75 samples? &nbsp;In other words, is Frame 1 centered on the 1st sample with +/- 75 samples? &nbsp;THis is what I would expect for this type of system, but it would violate the letter of the &quot;No Future Data&quot; rule. &nbsp;Is a fixed 75 sample lag acceptable for us to fully cover the +/- 150ms window? &nbsp;</p>",
  "messages": [
    {
      "id": "83453",
      "postDate": "07/05/2015 01:37:37",
      "content": "<p>Hello!</p>\n<p>In this project we're collecting data at 500Hz. &nbsp;Each sample is 2ms. &nbsp;As I &nbsp;understand it, the&nbsp;detection window in the discussion was +/- 75 samples wide. &nbsp;Does this mean that I can run with a lag of 75 samples? &nbsp;In other words, is Frame 1 centered on the 1st sample with +/- 75 samples? &nbsp;THis is what I would expect for this type of system, but it would violate the letter of the &quot;No Future Data&quot; rule. &nbsp;Is a fixed 75 sample lag acceptable for us to fully cover the +/- 150ms window? &nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "83458",
      "postDate": "07/05/2015 02:54:41",
      "content": "<p>This means that each event is tagged in the data with a window a +/- 75ms around the actual occurence of the event. For example, the class 'handstart' correspond to +/- 75ms starting from the instant where the user start moving his hand.&nbsp;</p>\n<p>From the challenge point of view, It would have make no sense to ask contestants to predict an event with a precision of 1 time sample. instead, you are asked to predict what the events files give you, i.e. a time window around the actual event.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "83513",
      "postDate": "07/06/2015 13:13:21",
      "content": "<p>You should still adhere to the &quot;no future data&quot; rule.</p>",
      "rawMarkdown": "You should still adhere to the \"no future data\" rule.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 83458,
      "author_name": "alexandrebarachant",
      "author_url": "",
      "post_date": "07/05/2015 02:54:41",
      "content": "<p>This means that each event is tagged in the data with a window a +/- 75ms around the actual occurence of the event. For example, the class 'handstart' correspond to +/- 75ms starting from the instant where the user start moving his hand.&nbsp;</p>\n<p>From the challenge point of view, It would have make no sense to ask contestants to predict an event with a precision of 1 time sample. instead, you are asked to predict what the events files give you, i.e. a time window around the actual event.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 83513,
      "author_name": "wcukierski",
      "author_url": "",
      "post_date": "07/06/2015 13:13:21",
      "content": "<p>You should still adhere to the &quot;no future data&quot; rule.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "83453": "",
    "83458": "",
    "83513": "You should still adhere to the \"no future data\" rule."
  },
  "source": "meta"
}