In the inline question 1 of the svm.ipynb workbook, the question asks:
"it is possible that once in a while a dimension in the gradcheck will not match exactly. What could such a discrepancy be caused by? Is it a reason for concern? What is a simple example in one dimension where a gradient check could fail? How would change the margin affect of the frequency of this happening? Hint: the SVM loss function is not strictly speaking differentiable"
Bolded sentence should probably be: How would changing the margin affect the frequency of this happening.
In the
inline question 1
of thesvm.ipynb
workbook, the question asks: "it is possible that once in a while a dimension in the gradcheck will not match exactly. What could such a discrepancy be caused by? Is it a reason for concern? What is a simple example in one dimension where a gradient check could fail? How would change the margin affect of the frequency of this happening? Hint: the SVM loss function is not strictly speaking differentiable" Bolded sentence should probably be: How would changing the margin affect the frequency of this happening.