BSN Contest Frequently Asked QuestionQ: Why should I participate?
A: The contest aims to introduce the BSN community to the data sharing idea. By participating you will:
Q: Does the competition have any research value for me?
Q: Please specify the tasks that will need to be addressed during the contest.
A: - This will be annoucned shortly.
Q: Data sets do not have explicit label attached to them. How should we train our system?
A: The details will be annoucned shortly.
Q: Can you specify the testing machine setup? What do you mean, when you ask for the file to not require additional libraries?
A: We will run the code on a 32 or 64 bit Windows machine with MATLAB, JAVA, and C++. By saying that it should be able to run without any additional information, we mean to say that all the libraries should be included in the executable.
Q: Can we run the final code on our machines?
A: Absolutely! Understand that we will have to take all measures to ensure the competition will be fair.
Q: According to the rules, the contest will be judged on both accuracy. Will you provide an objective function so that solutions can be optimized against it?
A: The objective function will be only based on the accuracy for now. That is, the how accurately the type of movements and the timing information will be determined. The exact objective function will be released in the coming weeks, certainly before April 1st.
Q: During the contest itself (once submissions have been received on your end), what will be the input provided to the algorithms and what should be the expected output?
A: During the contest, the input will be a movement file in one format. The output will be an answer to one of the 2 tasks listed above. We will provide the exact details of the tasks and the objective function shortly so you can start testing your system with respect to the exact criterion.
Q: What is the deadline for sending you the executables?
A: The exact time is to be determined yet. But it will be on May 10th.
Q: I noticed that for some pair of (movement,subject) the length of the data stream changes slightly for each mote. is this possible or is it an error in the data set?
A: The length of the data is affected by the way the data collection is stopped. While motes communicate all of their data concurrently, the incoming data is processed serially at the base station. What can happen is:
This only affects the tail of the data collection, and should not be a problem since we make sure to allow a bit of time after the last movement and before the data collection is stopped. Additionally, during the competition the movements have to be recognized within a 1000ms window centered around the point, which would more than compensate for this issue.