Looking over post on data description it looks like the "1" marker is related to the TrainLabels.csv. For example with the data for subject 6 session 01 (file Data_S06_Sess03) there's 60 "1"s in the list which corresponds to the number of entries in the TrainLabels.csv file:
\ngrep S06_Sess01 TrainLabels.csv | wc -l
\nshows 60. Looking at all the feedback numbers they are from 1 to 60. So far so good.
\nFor a FeedBackEvent array of 001010010000 does that mean that the letter flashed on at the 3rd entry? Then a time slice goes by with the first letter and then the next letter is flashed?
\nThat is up until the first 1 is seen in FeedBackEvent the test subject is sitting there waiting for the experiment to start.
", "messages": [ { "id": "61851", "postDate": "12/25/2014 03:12:47", "content": "Looking over post on data description it looks like the "1" marker is related to the TrainLabels.csv. For example with the data for subject 6 session 01 (file Data_S06_Sess03) there's 60 "1"s in the list which corresponds to the number of entries in the TrainLabels.csv file:
\ngrep S06_Sess01 TrainLabels.csv | wc -l
\nshows 60. Looking at all the feedback numbers they are from 1 to 60. So far so good.
\nFor a FeedBackEvent array of 001010010000 does that mean that the letter flashed on at the 3rd entry? Then a time slice goes by with the first letter and then the next letter is flashed?
\nThat is up until the first 1 is seen in FeedBackEvent the test subject is sitting there waiting for the experiment to start.
", "rawMarkdown": "", "votes": null }, { "id": "61855", "postDate": "12/25/2014 04:39:35", "content": "So from what I understand the test goes like this:
\nThe subject is asked to spell a certain word, one letter at a time. So for the word "pizza" they would first concentrate on the letter "P". After a certain amount of time the EEG signals are used in a classifier which then predicts which letter the subject was thinking of. It then displays the predicted letter on the screen, and this is the feedback event. The time when the prediction starts being displayed on the screen is marked with a 1 in the FeedBackEvent column in the dataset. The prediction is displayed on the screen for 1.3 seconds, after which the subject moves on to the next letter.
\nphalaris used this information in his Beat the Benchmark solution by training a gradient boosting classifier on the EEG signals for 1.3 seconds after each feedback event. You can check out the code here: http://www.kaggle.com/c/inria-bci-challenge/forums/t/11056/beating-the-benchmark-with-gbm
\nHope that helps!
", "rawMarkdown": "", "votes": null } ], "comments": [ { "id": 61855, "author_name": "tdevries", "author_url": "", "post_date": "12/25/2014 04:39:35", "content": "So from what I understand the test goes like this:
\nThe subject is asked to spell a certain word, one letter at a time. So for the word "pizza" they would first concentrate on the letter "P". After a certain amount of time the EEG signals are used in a classifier which then predicts which letter the subject was thinking of. It then displays the predicted letter on the screen, and this is the feedback event. The time when the prediction starts being displayed on the screen is marked with a 1 in the FeedBackEvent column in the dataset. The prediction is displayed on the screen for 1.3 seconds, after which the subject moves on to the next letter.
\nphalaris used this information in his Beat the Benchmark solution by training a gradient boosting classifier on the EEG signals for 1.3 seconds after each feedback event. You can check out the code here: http://www.kaggle.com/c/inria-bci-challenge/forums/t/11056/beating-the-benchmark-with-gbm
\nHope that helps!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "61851": "", "61855": "" }, "source": "meta" }