davidrosenberg/mlcourse

Undefined notation / notation bug in Lecture 1 Concept Check, Topic 2, Question 1

Opened this issue · 3 comments

davidrosenberg commented

What is N? If it's the minibatch size, it should be the subscript of m rather than n. @brett1479 right?

https://davidrosenberg.github.io/mlcourse/ConceptChecks/1-Lec-Check_sol.pdf

I think should be the same as: https://davidrosenberg.github.io/mlcourse/Archive/2017Fall/Lectures/02b.SGD.pdf#page=20
?

brett1479 commented

I think the issue here is that it should say that the full training set is $(x_1,y_1),\ldots,(x_N,y_N)$, since otherwise the mini-batch and the training set have the same size (which is what it says now). We can swap them to make them agree with the lecture.

venik commented

It would make sense to add 1 line with P(m_1 = i) under the sum, to make it a bit more clear from where \frac{1}{n} came from.
https://davidrosenberg.github.io/mlcourse/Archive/2017Fall/Lectures/02b.SGD.pdf#page=20