Quote:
Originally Posted by rita.wu
Dear Prof, I've implemented the PLA algorithm with Matlab, and submitted my answers. However, I find that deciding the iteration of 10 samples is quite troublesome. As in my implementation, even after repeating the experiment for 1000 times, the answer is just between two choices you provided, sometimes closer to the first one, while sometimes the second.
Besides, I changed the way of picking the misclassified point in each iteration, the iteration numbers changed, too. So I think the answer depends on the implementation of PLA and the generation of samples points.
Could you please shed light on the problem? Thank you~

There are variations depending on the factors you mentioned. However, if you repeat the experiment for a large number of runs (different targets, different data, etc.), and
take the average as specified in the problem statement, you should be closer to one of the given answers. Is this what you have done?