Using same sample for multiple times on Perceptron training

[Sam Zhang]

In problem 2.2, my understanding of the problem is we have a sample size of m, yet going through the m samples one by one, every time updating the weight vector once is not guaranteed to return a model that correctly classify all the m samples and we have to run more up to 2^m times. Does that mean it's ok to update a perceptron model using the same sample more than once, and even consecutively using it?

Thanks,
Sam 

[Mu Li]

Hi Sam,

You are right. The sequence of samples, typically quite long, given to perceptron can be constructed from repeating and reordering  a smaller sample set. In fact we often do that on practice. 

Thanks,

Mu

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-201...@googlegroups.com.
To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.